Biologically Inspired Feature Extraction for Human Face Recognition using Multi Angle Movement

Human Face Recognition plays an important role in many applications. In the past, many researchers tend to use cluster propagation mechanism and wavelet model as data representation for Human face recognition. As different visual face feature of an unconstrained video describes different aspects about motion data and they have dissimilar discriminative power, all the features in the template were not extracted, therefore compromising retrieval performance. In this paper, a new method called, Temporal Average Multi-angle Movement Feature Selection (TA-MMFS) based on key-frame selection is developed for human face recognition which identifies the frames carrying the most discriminatory motion features. TA-MMFS initially removes the unwanted noise by applying Temporal Average Filtering-based Pre-processor. With the de-noised preprocessed image, TA-MMFS combines optical flow with biologically inspired features of face to extract new feature descriptor using Multi-angle Movement Feature Selection. The TA-MMFS therefore extracts the most discriminatory motion features with respect to time which is informative for human face recognition. It achieves a maximum feature extraction efficiency gain of 18% and 18% minimizing the feature extraction time. The experimental results demonstrate that the new method is more efficient than reference methods in terms of noise removal. Extensive experiments demonstrate the effectiveness of the proposed TA-MMFS over the state-of-art methods for human face recognition.