ECG classification system based on multi-domain features approach coupled with least square support vector machine...

Area of Science:

• Biometrics and Security Engineering
• Signal Processing and Machine Learning

Background:

• Traditional password authentication methods exhibit significant weaknesses in speed and integrity.
• Biometric authentication, particularly using electrocardiogram (ECG) signals, is gaining prominence due to its unique and difficult-to-counter nature.

Purpose of the Study:

• To develop a robust ECG-based authentication system.
• To investigate the effectiveness of multi-domain features for ECG verification.
• To compare the performance of different classifiers for ECG authentication.

Main Methods:

• Extraction of time and frequency domain features from ECG signals using an optimized Triple Band filter bank.
• Feature selection to identify the most relevant and non-redundant features.
• Classification using Least Square Support Vector Machine (LS-SVM), K-means, and K-nearest algorithms.

Main Results:

• The proposed ECG biometric authentication system demonstrated superior performance compared to existing methods.
• Individual feature domains achieved accuracies of 88% (time) and 95% (frequency).
• A combination of time and frequency features yielded an outstanding accuracy of 99%.

Conclusions:

• The proposed multi-domain feature extraction coupled with LS-SVM provides a highly accurate and robust ECG biometric authentication solution.
• Combining time and frequency domain features significantly enhances the performance of ECG-based verification systems.
• The developed model shows great potential for securing device data through reliable biometric identification.