Accurate detection of human faces in static or video images, is the basis in many applications such as face recognition, human-machine interface system, tracking and video surveillance, model-based video coding and security access control. There are many challenges for intelligent proposed systems that face detection is one of them. Detection rate and number of false detection are used for comparisons with face detection systems. The purpose of this research is to provide a face detection system which is based on wavelet network. The density of each image window first is equalized for reduction of illumination effects and PCA is used for dimensionality reduction and feature extraction. This step is extremely effective to streamline the complexity of the system. Then, wavelet network is used for classification. This wavelet network is a member of Fixed Grid Wavelet Network that is formed with no need of training. Another goal of this research is to compare the neural network with wavelet network. To get this aim, wavelet network has been replaced with neural network and the results had been compared with each other. To reduce computation of the hardware implementation on FPGA, input and output parameters of wavelet function are computed and stored in a LUT that the norm of input vector is connected to its address line. To calculate the norm of input vector, a proposed structure, based on the parallel squares units and Carry Save Adders (CSA), is used. After calculation of product phrase, Wallace Tree Adder (WTA) is used to sum the products in square units. The proposed structure has a low power consumption and short hardware data path. The simulation results have demonstrated the good performance of the structure. Keywords : face detection, principal component analysis, neural networks, wavelet network
