Abstract
Face is a complex multidimensional visual model and it is difficult to develop a computational model for recognition. A novel approach is presented to face recognition in this paper, which uses wavelet transform (WT), fast independent component analysis (FastICA) and radial basis function (RBF) neural networks. Firstly, low frequency subband images are extracted from original face image by 2D wavelet transform. Secondly, for reducing computational cost and converges difficultly, improved FastICA is applied to extract features from the low frequency subband image. Then, the extracted features are classified through RBF neural networks. Lastly, the proposed algorithm is tested on the ORL face database and result shows that it has good performance both in terms of recognition accuracy and robustness.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call