Development of a biometric authentication system based on HAAR transformation and Score Level Fusion

In traditional authentication system Password, Pin-number and Signature are used as sources for identification. The latest technological advancements forced human beings towards the development of complex authentication systems based on biometric traits. The most challenging tasks in biometric authentication system are the recognition of an accurate matching biometric trait in the database and also the size of database to be searched. In this paper, a bimodal biometric authentication approach is introduced where two biometric traits namely facial expressions and fingerprints are used. In the proposed method, Discrete HAAR wavelet compression, HOG (Histogram of Oriented Gradients) for feature extraction and GMM (Gaussian Mixer Model) are applied sequentially on both facial and finger print datasets to derive recognition templates of each trait category. Finally respective recognition templates of each category are fused together using score level fusion to get composite recognition template. Experiments conducted on both finger and face data of 10 individuals. It is observed that proposed model exhibits good accuracy rates and the model helps to reduce the size of database and to retrieve the data from the databases for effective recognition with minimum timestamp.