Face Recognition Based on FastICA and RBF Neural Networks

The face is a complex multidimensional visual model and it is difficult for face recognition to develop a computational model. This paper, a novel approach is presented to face recognition, which combines Fast independent component analysis (FastICA) and Radial Basis Function neural networks. Firstly, in order to reduce the image data, low-frequency subband images are extracted from original face images by 2D wavelet transform. After, FastICA is applied to extract features from the low-frequency subband image which contains most discriminated information of face image. For reducing computational cost, the improved FastICA method is introduced. Then, RBF neural networks classifier is designed. Lastly, this algorithm is tested on the ORL face databases and the experimental results show that the method has good performance in terms of recognition accuracy and the robustness is enhanced greatly.