(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Development of heartbeat based biometric system

Development of heartbeat based biometric system / Chin Chee Yeen
Satu dekad yang lalu, penggunaan isyarat elektrokardiogram (ECG) dalam biometrik telah meningkat. Data isyarat ECG telah dieksploitasi sebagai biometrik untuk identifikasi dan juga klasifikasi individu adalah disebabkan oleh ciri-ciri uniknya. ECG merekod aktiviti jantung manusia dan ECG adalah unik kerana setiap individu mempunyai struktur jantung yang berbeza antara seorang individu dengan individu yang lain. Oleh sebab itu, ciri-ciri ECG adalah berbeza antara individu dan ECG mempunyai infomasi penting untuk membezakan seorang individu itu daripada individu lain. Informasi sistem pengesahan biometrik yang sedia ada seperti cap jari, pengecaman muka dan suara boleh dipalsukan. Penggunaan ECG amat sesuai untuk pengesahan biometrik kerana ECG tidak mungkin dapat dipalsukan. Projek ini mencadangkan pendekatan klasifikasi dan analisis ECG menggunakan Penjelmaan Wavelet yang merupakankaedah untuk analisis masa-frekuensi dan isyarat pengekodan serta Support Vector Machine (SVM). Pendekatan ini termasuk empat fasa, iaitu pra-pemprosesan, pembahagian, pengekstrakan ciri-ciri serta klasifikasi isyarat. Penjelmaan Wavelet digunakan untuk pemprosesan dan penyahhingaran isyarat. Selepas pembahagian, Wavelet ibu “db4”, “sym3” dan “coif2” digunakan untuk mengekstrak ciri-ciri ECG. Mereka akan digunakan sebagai ciri input dalam fasa klasifikasi mana SVM akan diaplikasikan. Data ECG diperolehi daripada pangkalan data atas talian, iaitu pangkalan data Physionet ECG-ID. Sistem biometrik ini dinilaikan dengan menggunakan 298 rakamanECG daripada 79 individu. Antara 90 individu tersebut, terdapat 44 orang lelaki dan 46 orang wanita yang berumur dari 13 tahun sehinggan 75 tahun. Bilangan rakaman ECG bagi setiap individu adalah sebanyak 2 hinggan 20. Setiap rakaman dirakamkan sepanjang 20 saat, didigitasi dengan 500Hz dan berdasarkan ECG 1-plumbum. Prestatsi sistem dinilai dengan menggunakan tahap penyahhingaran, kaedah ambang dan jenis wavelet ibu yang berbeza. Sistem gabungan tahap pengyahhingaran 3, keadah ambang “heursure” dan wavelet ibu “db6” menghasilkan prestasi yang terbaik, iaitu 92.5% GAR bagi 5% FAR dan 6.9499% EER. _______________________________________________________________________________________________________ During last decade, the use of electrocardiogram (ECG) signal in biometric recognition has increased. ECG signal data is exploited as biometrics for human identification as well as classification due to its unique characteristics. The ECG records the heart activity and it is unique as different individualshave distinct heart structure. Hence, ECG characteristics are distinct among individuals and ECG provides vital information to differentiate one individual from another. The existing biometric authentication systems such as fingerprint, facial or voice are not reliable as the information can be counterfeited. The made ECG is suitable for human recognition as it is impossible to be faked.This project proposes a reliable partially fiducial ECG analysis and classification approach using wavelet transform, a powerful time-frequency analysis and signal coding tool and Support Vector Machine classifier. The approach includes four phases, which are signal preprocessing, segmentation, feature extraction and classification. Wavelet transform is employed for signal preprocessing and also de-noising. After segmentation, the coefficient of transform of ECG signals are extracted using mother wavelets “sym3”, “db4” and “coif2”. They are then used as the input features in the classification phase where Support Vector Machine (SVM) is used as the classifier. The ECG database is obtained from the publicly available database, Physionet ECG-ID database. The number of subjects involved in this project is 79 persons, total of 298 ECG recordings. The ECG records were obtained from 44 men and 46 women aged from 13 to 75 year old. The number of records for each subject varies from 2 to 20 and each recording contains ECG lead 1, recorded for 20 seconds and digitized at 500Hz, which will be used in the classification phase. The system performance was evaluated by varying parameters including wavelet de-noising order, threshold method and type of mother wavelet. The combination of de-noising order 3, “heursure” threshold method and “db6” mother wavelet yield the top performance of GAR of 92.5% at FAR of 5%, and EER of 6.9499%.
Contributor(s):
Chin Chee Yeen - Author
Primary Item Type:
Final Year Project
Identifiers:
Barcode : 00003107120
Accession Number : 875007240
Language:
English
Subject Keywords:
electrocardiogram (ECG) signal; biometric recognition; unique characteristics
First presented to the public:
6/1/2017
Original Publication Date:
4/16/2018
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 69
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2018-04-16 16:54:47.506
Date Last Updated
2019-01-07 11:24:32.9118
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Development of heartbeat based biometric system1 2018-04-16 16:54:47.506