BPNN-hippoamy Algorithm for Statistical Features Classification

Neural Network is the simplified model of the biological nervous system, due to which several well defined architectures of Neural Network have been developed. But some factors affect the performance of the Neural Network such as number of training, number of neurons. In order to increase the performance of network by minimizing all these factors, a modified new algorithm “Back Propagation Neural Network- Hippoamy algorithm” is proposed based on the concept of architecture re-usability for face image classification. The proposed algorithm is experimentally tested on IMM frontal face database which consists of 240 sample images of 40 different persons and these samples are analyzed using statistical features like maximum probability, contrast, correlation, energy, homogeneity and entropy of gray level co-occurrence matrix. The proposed algorithm is compared with the conventional Back Propagation Neural Network and the results of performance metrics – acceptance ratio, mean square error, suggest that the modified algorithm minimizes all these factors and is well suited for face classification and recognition.