A Novel BP Neural Network Based System for Face Detection

We describe a new neural network, which can improve the performance of face detection system. In this paper, we propose a system that combines the Gabor feature and momentum factor back propagation algorithm for face detection. First, the Gabor feature of the training set is extracted and is inputted to the momentum factor of Back Propagation neural network for training. Then, using the trained system detects whether the face targets exist in the input image, and marking the target with the window. In order to enhance the training effect of the traditional Back Propagation neural network, the momentum factor is added to the Back Propagation algorithm, which can effectively slow down the trend of the network training in the shock and avoid the algorithm drop into the local minimum. Furthermore, the added momentum factor can adaptively adjust each layer weight of the Back Propagation neural network. Extensive experimental results demonstrate that our solution is effective and also competitive, compared to the classic and also state-of-the-art face detection models.