MULTIMODAL HYBRID BIOMETRIC IDENTIFICATION USING FACIAL AND ELECTROCARDIOGRAM FEATURES

Abstract

A single biometric is always prone to errors and misleading results. The inclination has been towards employing two or more biometric traits for designing any schema, accomplishing a superior efficiency. As per the application, these schemas can be deployed for identification as well for recognition, proving to be instrumental in multitudinal fields. Biometric traits examined in the proposed schema are electrocardiogram and face traits. For the face traits, endeavour has been put for entropy biased identification, employing DCT before PPCA. Further exploration of Electrocardiogram signal has been compassed, including the diverse feature points and classifiers. Fiducial points and temporal locale of these points, combined with entire PQRST segment, prove to be an enriching feature set, resulting in much improved results. The score level fusion is exercised along with normalization. The novel fusion schema worked for these two biometric traits does not put any bias on any the traits but works to add the end outputs of the two separate schemas, such as they are employed independently, the best of which is selected as end identification of personal. The schema is certified using the Yale facial dataset and ECG-ID set from Physionet. This addition has shown tremendous enrichment to our accuracy results, strengthening our methodology.