Research Of Face Image Recogntion Based On Wavelet Neural Networks

Human face image recognition is a research focus of artificial intelligence and computer vision. Human face information could be obtained from non-contact method, such as cameras. Therefore it is very suitable for identification authentication. In the modern society, a new type of security system is urgently needed. This system has to protect the safety of people’s lives and properties. Meanwhile, it should not affect people’s daily life. Human face image recognition can meet these demands perfectly, so it is widely used.This paper studies the basic algorithm of wavelet neural networks, and applies it to face image recognition. The algorithm’s main idea is to obtain wavelet neural network algorithm, combining wavelet transform theory and neural network theory. In the process of face image recognition experiments, the adaptability, validity, convergence and correctness of wavelet neural networks are verified. Furthermore, this paper compares the application to face image recognition of three different kinds of wavelet neural networks. After the simulating experiments, PNN is proved to possess better recognition accuracy and real-time ability. Therefore, PNN can be applied to real-time signal processing. In the experiments, this paper adopts ORL human face database.This paper contains the main contents as follows:(1) Introduce the application prospect and research status of face image recognition technology. Present the principle, property and characteristic of the related digital image process used in the experiments. Introduce and analyze the typical algorithms of wavelet technology and the algorithm of DCT respectively.(2) Study the algorithm of BP LVQ and PNN and their application to face image recognition. The simulating results show that the recognition accuracy, stability and real-time ability of PNN are better than the other two neural networks.(3) Analyze elaborately the algorithm of wavelet neural networks. Construct wavelet neural networks by choosing different wavelet base, and apply them to face image recognition.(4) At last, this paper simulates the algorithm of wavelet neural networks with the software MATLAB. The experimental results show that wavelet neural networks possess better recognition accuracy, convergence and stability. To some extent, wavelet neural networks satisfy the basic demands of face image recognition technology, and possess research, practical and commercial value.