Neural network based face recognition with moment invariants

This paper introduced an experimental evolution of the effectiveness of utilizing various moments as pattern features in human face technology. In this paper, we apply Pseudo Zernike Moments (PZM) for recognition human faces in two-dimensional images, and we compare their performance with other type of moments. The moments that we have used are Zemike Moments (ZM), Pseudo Zernike Moments (PZM) and Legendre Moments (LM). We have used shape information for human face localization, also we have used Radial Basis Function (RBF) neural network as classifier for this application. The performance of classification is dependent on the moment order as well as the type of moment invariant, but the classification error rate was below %10 in all cases. Simulation results on face database of Olivetti Research Laboratory (ORL) indicate that high order degree of Pseudo Zemike Moments contain very useful information about face recognition process, while low order degree contain information about face expression. The PZM of order of 6 to 8 with %1.3 error rate are very good features for human face recognition that we have proposed.