Setisi 2015 best paper

24
Biometrik Detak Jantung berdasarkan Sinyal Photoplethysmography I Ketut Edi Purnama 1 , Mauridhi Hery Purnomo 2 , Shi-Jinn Horng 3 , Raudhatul Jannah 4 , Fakarudin Afdlol 5 1,2,4 Institut Teknologi Sepuluh Nopember 3 National Taiwan University of Science and Technology 5 Politeknik Elektronika Negeri Surabaya

Transcript of Setisi 2015 best paper

Page 1: Setisi 2015 best paper

Biometrik Detak Jantungberdasarkan Sinyal

PhotoplethysmographyI Ketut Edi Purnama 1 Mauridhi Hery Purnomo 2 Shi-Jinn Horng 3

Raudhatul Jannah 4 Fakarudin Afdlol 5

124Institut Teknologi Sepuluh Nopember3National Taiwan University of Science and Technology

5Politeknik Elektronika Negeri Surabaya

Outline

bull Pendahuluan

bull Penelitian sebelumnya

bull Metodebull Pengambilan data

bull Segmentasi dan Fitur ekstraksi

bull Klasifikasi

bull Hasil

bull kesimpulan

Pendahuluan

bull Identifikasi Biometric fingerprint face detection iris hand geometry

bull Sistem keamanan

bull Perangkat PPG menerima dan merespon suatu sinyal atau stimulus pada teknologi pulse

oximeter untuk menangkap perubahan volume darah berdasarkan Light Emitting Dioda

(LED) Karena perubahan jumlah volume darah sesuai (sinkron) dengan detak jantung

teknik PPG dapat digunakan untuk mengukur kecepatan detak jantung

Penelitian sebelumnya

bull Y N Singh dan P Gupta

menggunakan matching decision untuk mencari korelasi terhadap fitur yang

dimiliki sinyal detak jantung ECG pada setiap individu

bull P Spachos J Gao dan D Hatzinakos

data diambil menggunakan Pulse Oximeter Sensor hasil menunjukkan pada subjek

yang sama memiliki time interval yang konstan secara statistic terhadapmaksimumminimum poin yang dihasilkan

bull J Yao X Sun dan Y Wan

Data yang dipakai menggunakan BvpPLUX System dari OpenSignal PPG Dataset

serta menggunakan NONIN pulse oximeter dari BioSec PPG Dataset Hasil

identifikasi tergantung dari dataset yang digunakan dimana jika sinyal dataset

tidak stabil maka tingkat akurasi lebih rendah Hal ini dapat terjadi karena

pengaruh dari rangkaian sensor maupun kondisi saat pengambilan data

Pengklasifikasian diaplikasikan menggunakan Nearest neighbor and majority voting

untuk mencocokkan data sinyal input

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 2: Setisi 2015 best paper

Outline

bull Pendahuluan

bull Penelitian sebelumnya

bull Metodebull Pengambilan data

bull Segmentasi dan Fitur ekstraksi

bull Klasifikasi

bull Hasil

bull kesimpulan

Pendahuluan

bull Identifikasi Biometric fingerprint face detection iris hand geometry

bull Sistem keamanan

bull Perangkat PPG menerima dan merespon suatu sinyal atau stimulus pada teknologi pulse

oximeter untuk menangkap perubahan volume darah berdasarkan Light Emitting Dioda

(LED) Karena perubahan jumlah volume darah sesuai (sinkron) dengan detak jantung

teknik PPG dapat digunakan untuk mengukur kecepatan detak jantung

Penelitian sebelumnya

bull Y N Singh dan P Gupta

menggunakan matching decision untuk mencari korelasi terhadap fitur yang

dimiliki sinyal detak jantung ECG pada setiap individu

bull P Spachos J Gao dan D Hatzinakos

data diambil menggunakan Pulse Oximeter Sensor hasil menunjukkan pada subjek

yang sama memiliki time interval yang konstan secara statistic terhadapmaksimumminimum poin yang dihasilkan

bull J Yao X Sun dan Y Wan

Data yang dipakai menggunakan BvpPLUX System dari OpenSignal PPG Dataset

serta menggunakan NONIN pulse oximeter dari BioSec PPG Dataset Hasil

identifikasi tergantung dari dataset yang digunakan dimana jika sinyal dataset

tidak stabil maka tingkat akurasi lebih rendah Hal ini dapat terjadi karena

pengaruh dari rangkaian sensor maupun kondisi saat pengambilan data

Pengklasifikasian diaplikasikan menggunakan Nearest neighbor and majority voting

untuk mencocokkan data sinyal input

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 3: Setisi 2015 best paper

Pendahuluan

bull Identifikasi Biometric fingerprint face detection iris hand geometry

bull Sistem keamanan

bull Perangkat PPG menerima dan merespon suatu sinyal atau stimulus pada teknologi pulse

oximeter untuk menangkap perubahan volume darah berdasarkan Light Emitting Dioda

(LED) Karena perubahan jumlah volume darah sesuai (sinkron) dengan detak jantung

teknik PPG dapat digunakan untuk mengukur kecepatan detak jantung

Penelitian sebelumnya

bull Y N Singh dan P Gupta

menggunakan matching decision untuk mencari korelasi terhadap fitur yang

dimiliki sinyal detak jantung ECG pada setiap individu

bull P Spachos J Gao dan D Hatzinakos

data diambil menggunakan Pulse Oximeter Sensor hasil menunjukkan pada subjek

yang sama memiliki time interval yang konstan secara statistic terhadapmaksimumminimum poin yang dihasilkan

bull J Yao X Sun dan Y Wan

Data yang dipakai menggunakan BvpPLUX System dari OpenSignal PPG Dataset

serta menggunakan NONIN pulse oximeter dari BioSec PPG Dataset Hasil

identifikasi tergantung dari dataset yang digunakan dimana jika sinyal dataset

tidak stabil maka tingkat akurasi lebih rendah Hal ini dapat terjadi karena

pengaruh dari rangkaian sensor maupun kondisi saat pengambilan data

Pengklasifikasian diaplikasikan menggunakan Nearest neighbor and majority voting

untuk mencocokkan data sinyal input

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 4: Setisi 2015 best paper

Penelitian sebelumnya

bull Y N Singh dan P Gupta

menggunakan matching decision untuk mencari korelasi terhadap fitur yang

dimiliki sinyal detak jantung ECG pada setiap individu

bull P Spachos J Gao dan D Hatzinakos

data diambil menggunakan Pulse Oximeter Sensor hasil menunjukkan pada subjek

yang sama memiliki time interval yang konstan secara statistic terhadapmaksimumminimum poin yang dihasilkan

bull J Yao X Sun dan Y Wan

Data yang dipakai menggunakan BvpPLUX System dari OpenSignal PPG Dataset

serta menggunakan NONIN pulse oximeter dari BioSec PPG Dataset Hasil

identifikasi tergantung dari dataset yang digunakan dimana jika sinyal dataset

tidak stabil maka tingkat akurasi lebih rendah Hal ini dapat terjadi karena

pengaruh dari rangkaian sensor maupun kondisi saat pengambilan data

Pengklasifikasian diaplikasikan menggunakan Nearest neighbor and majority voting

untuk mencocokkan data sinyal input

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 5: Setisi 2015 best paper

bull J Yao X Sun dan Y Wan

Data yang dipakai menggunakan BvpPLUX System dari OpenSignal PPG Dataset

serta menggunakan NONIN pulse oximeter dari BioSec PPG Dataset Hasil

identifikasi tergantung dari dataset yang digunakan dimana jika sinyal dataset

tidak stabil maka tingkat akurasi lebih rendah Hal ini dapat terjadi karena

pengaruh dari rangkaian sensor maupun kondisi saat pengambilan data

Pengklasifikasian diaplikasikan menggunakan Nearest neighbor and majority voting

untuk mencocokkan data sinyal input

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 6: Setisi 2015 best paper

bull R Kavsaoğlu A K Polat dan M R Bozkurt

Data diambil menggunakan mikrokontroler dan sensor DCM03 serta sistem klasifikasi

menggunakan metode k-nn (k-Nearest Neighbor) Lima fitur berbeda digunakan

untuk tahap ekstraksi fitur seperti augmentation index puncak sistolik dan diastolik

lebar pulsa dan peak-to-peak interval

bull Y Y Gu dan Y T Zhang

Proses ekstraksi fitur menggunakan empat fitur yang berbeda seperti upward slope

downward slope dan time interval Fuzzy logic diaplikasikan melalui decision

making untuk human verification

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 7: Setisi 2015 best paper

Pengambilan Data

Rangkaian arduino dan Pulse Sensor

untuk pengambilan data

Bentuk sinyal PPG yang dihasilkan oleh rangkaian

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 8: Setisi 2015 best paper

Blok Diagram Sistem Identifikasi Detak JantungSinyal PPG

Sinyal Input

Segmentasi

Fitur EkstraksiIdentifikasi

(Klasifikasi) DetakJantung

Hasil

(Detak jantungteridentifikasi)

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 9: Setisi 2015 best paper

Sinyal

PhotopletysmographyData Latih Data Uji

Data penelitian

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 10: Setisi 2015 best paper

Blok Diagram Proses Pengembangan AlgoritmaIdentifikasi Detak Jantung PPG (I)

Segmentasi danFitur Ekstraksi

Data sinyal PPG Segmentasi sinyal

Fitur ekstraksi

(deteksi nilaipuncak + LDA)

A

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 11: Setisi 2015 best paper

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600

Am

plit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil fitur ektraksi pada tiga sampel data sinyal PPG

Hasil Segmentasi Sinyal dan Ektraksi Fitur

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 12: Setisi 2015 best paper

Metode Linear Discriminant Analysis

x

y

K+[xy] L+sum(([xy] Q) [xy]2) = 0

100 200 300 400 500 600 700 800 900

450

500

550

600

650

Metode Linear Discriminant Analysis (LDA)

digunakan untuk mendapatkan nilai fitur

yang didapatkan dari proses ekstraksi fitur

kemudian mengklasifikasikannya sesuai

dengan kelompoknya

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 13: Setisi 2015 best paper

Blok Diagram Proses Pengembangan Algoritma IdentifikasiDetak Jantung PPG (II)

Klasifikasi

Data Training

+

Data Testing

Naiumlve Bayes amp SMO

B

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

0 100 200 300 400 500 600 700 800 900 1000450

500

550

600A

mplit

ude

0 100 200 300 400 500 600 700 800 900 1000400

500

600

700

Sample Index

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 14: Setisi 2015 best paper

Hasil Klasifikasi

TP rate FP rate ACC()Naiumlve Bayes 078 0056 78048

SMO 0829 0042 8292

HASIL PROSES PENGUJIAN OUTPUT DENGAN TARGET

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 15: Setisi 2015 best paper

Pengujian Data Clustering

500 520 540 560 580 600 620 640 660 680410

420

430

440

450

460

470

480

490

500

Nilai P

uncak M

inim

um

Nilai Puncak Maksimum

subject1

subject2

subject3

subject4

subject5

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 16: Setisi 2015 best paper

Kesimpulan

bull Sistem yang dirancang dapat mengidentifikasi detak jantung dari masing-masing

individu

bull Ekstraksi ciri berdasarkan nilai puncak maksimum dan minimum dari sinyal

Photoplethysmograph

bull Proses identifikasi dikembangkan dengan metode Naiumlve Bayes dan SMO Akurasi

yang baik dibuktikan dalam identifikasi detak jantung terhadap 10 buah sinyal uji

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 17: Setisi 2015 best paper

Penelitian Selanjutnya

bull Pada penelitian selanjutnya dibutuhkan ekstraksi fitur yang lebih kompleks agar

pengenalan ciri sinyal Photoplethysmograph dari setiap individu semakin mudah untuk

dibedakan

bull Fitur ekstraksi tersebut dapat berupa nilai time interval tinggi puncak sistolik tinggi puncak

diastolik dan jarak dari puncak sistolik ke puncak diastolik

bull Data subjek yang dijadikan sampel juga perlu diperbanyak untuk pengujian sistem

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 18: Setisi 2015 best paper

Referensi

bull peakdet Peak detection using MATLAB ldquohttpwwwbillauercoilpeakdethtmlrdquo

bull P Spachos J Gao dan D Hatzinakos ldquoFeasibility study of photoplethysmographic signals for biometric identificationrdquo in Proc of the 17th Int Conf on Digital Signal Processing (DSP) 2011 pp 1 ndash 5

bull J Yao X Sun dan Y Wan ldquoA pilot study on using derivatives of photoplethysmographic signals as a biometric identifierrdquo in Proc of the 29th IEEE Annual Int Confof the Engineering in Medicine and Biology Society (EMBS) 2007 pp 4576 ndash 4579

bull Y N Singh dan P Gupta ldquoCorrelation-based classification of heartbeats for individual identificationrdquo Soft Computing vol 15 no 3 pp 449ndash460 2013

bull M Joel dan G Yury ldquohttppulsesensorcomrdquo

bull Biel L Pettersson O Lennart P dan Peter W (2001) ECG analysis a new approach in human identification IEEE Trans Instrum Meas 50(3)808ndash812

bull Li C Zheng C dan Tai C (1995) Detection of ECG characteristics points using wavelet transforms IEEE Trans Biomed Eng 42(1)21ndash28

bull R Kavsaoğlu A K Polat dan M R Bozkurt ldquoA Novel Feature Rangking Algorithm for Biometric Recognition with PPG Signalsrdquo Computers in Biology and Medicine 49 (2014) 1-14

bull Y Y Gu dan Y T Zhang Photoplethysmographic authentication through fuzzy logic in IEEEE MBS Asian-Pacific Conference on Biomedical Engineering 20ndash22 Oktober 2003 pp 136ndash137

bull Zhu M (2001) rdquoFeature Extraction and Dimension Reduction with Applications to Classification and the Analysis of Co-occurrence Datardquo doctoral dissertation Stanford University

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 19: Setisi 2015 best paper

TERIMA

KASIH

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 20: Setisi 2015 best paper

Skematik - PulseSensor

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 21: Setisi 2015 best paper

Have you considered the recognition accuracy if the subject did some activity as you know rest (sitting) and walking have a significance bpm difference

bull Berdasarkan penelitian ldquoHow accurate is pulse rate variability as an estimate of heart rate variabilityrdquo

bull Results speak in favor of sufficient accuracy when subjects are at rest although many studies suggest that short-term variability is somewhat overestimated by PRV which reflects coupling effects between respiration and the cardiovascular system Physical activity and some mental stressors seem to impair the agreement of PRV and HRV often to an inacceptable extent Findings regarding the position of the sensor or the detection algorithm are not conclusive

bull PRV as an estimate of HRV has been proved to be sufficiently accurate only for healthy (and mostly younger) subjects at rest

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 22: Setisi 2015 best paper

Heartbeat monitor needs around 5-10 seconds to calculate bpm (beat per minute) Seen from this articles that to recognize need 31 beat (shown in fig4) so it need almost half minute data (if the subject have 70bpm in rest condition) and it really not realtime systems Compare with fingerprint which is in order of miliseconds for recognition

bull Perlu dicoba

bull Data training 1 menit

bull Data testing 5 detik

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya

Page 23: Setisi 2015 best paper

In order to do an experiments for biometrics recognition author need to take more than 20 id this article using only 5 id

bull Tambah data sampelnya