Biometric analysis using fused feature set from side face texture and electrocardiogram

Abstract

Multimodal biometric authentication requires fusion of information extracted from different biometric modalities. Face recognition is the most common and versatile biometric parameter used for years. Recently, biosignals such as electrocardiogram (ECG), photoplathysmogram etc. are under study for probable use in authentication work. It is also established that multi-parameter approach in biometric analysis plays a vital role in increasing accuracy and robustness and preventing spoofing in spite of more computational demand. In the present study, information fusion based authentication system is proposed using face and ECG. Instead of conventional face image, a unique frontal face textural signal is proposed. This leads to simpler data processing similar to that of ECG signal. Finally, information from both the signals is fused at mother template generation level. A good accuracy is achieved using mean square deviation method as presented in the result section. A stability study is also made with five volunteers to check the long term variability of the features.