Cardioid graph based ECG biometric recognition incorporating physiological variability detection /
Security concerns regarding individual identification can be handled by utilizing unique biometric traits available on a person. Unlike traditional identification mechanisms which must either be carried with us or remembered, biometrics are characteristics of human beings. The electrocardiogram (ECG...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur :
Kulliyyah of Engineering, International Islamic University Malaysia,
2015
|
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4389 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Security concerns regarding individual identification can be handled by utilizing unique biometric traits available on a person. Unlike traditional identification mechanisms which must either be carried with us or remembered, biometrics are characteristics of human beings. The electrocardiogram (ECG) traces are unique to an individual due to different body shape, ionic composition and etc. However, the state of a human being's posture and activity level affects heartbeat rates, which in turn disturbs the regular ECG pattern. If a person's ECG traces during different activity levels are considered as the identification parameter, even though the ECG belongs to the same particular person, there will always be dissimilarity present between each recording. Since the feasibility and accuracy of automatic identification via ECG is higher when the variability rate or differences among ECG recordings are lower, these anomalies would hinder person identification. When considering potential projects for biometric identification via ECG at different physiological conditions, the enrolment population can be very large. For a project of such large scale, it is also necessary to think about the complexity of the system design, storage availability and system execution time. It is a normal belief, that complexity ensures safety, however complexity also means time delay. Therefore, in this study, feature extraction method based on Cardioid graph has been chosen and will be implemented, due to its simplicity and faster processing time. To the best of our knowledge, ECG biometric identification at varying physiological conditions based on Cardioid graph has never been reported. This issue will be the main focus of the study. The research is conducted on ECG data acquired from 30 subjects while they are carrying out six daily activities. An average identification rate of 96.4% has been achieved as compared to previous works. Furthermore, a novel model of compressed ECG biometric identification at varying physiological conditions based on Cardioid graph has also been proposed and experimentally verified. Not only will compressed ECG lead to reduced file size but also faster execution time. According to the results, when compared to our previous experiment with non-compressed ECG as the input, the same identification rate of 96.4% can be achieved. When the file size has been compressed up to 73.3%, the execution time is minimized by 63%. However, further higher identification rate of 96.7% can be achieved at the expense of lower compression ratio of 50% while at the same time reducing the execution time by 51.7%. Therefore, having an easy system that is capable of producing high identification rates with both compressed and non-compressed ECG signal in different physiological conditions promises a robust and reliable biometric identification system with wider range of implementation. A new assimilation of such a biometric identification system has been proposed that expects to have the abovementioned features as the input and at the same time, capable of performing identification accurately at the output. |
|---|---|
| Physical Description: | xvi, 85 leaves : ill. ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 74-79). |