A rotation invariant algorithm for recognition

Abstract

In this paper, a simple technique for human face classification using two transforms and neural nets is introduced. A new concept for rotation invariant based on Fourier descriptors and neural networks is presented. Although, the magnitude of the Fourier descriptors is translation invariant, there is no need for scaling or translation invariance. This is because the face sub-image here (20x20 pixels) is segmented from the whole image during the detection process as in my previous work [12]. The feature extraction algorithm based on Fourier descriptors is modified to reduce the number of neurons in the hidden layer. The second stage extracts wavelet coefficients of the resulted Fourier descriptors before application to neural network. Wavelet transforms have been shown to provide advantages in terms of better representation for a given data to be compressed. The final vector is fed to a neural net for face classification. Simulation results for the proposed algorithm show a good performance compared with previous results. © Springer-Verlag 2001.