Walsh Functions And Their Application To Human Face Recognition

As all knows, the Walsh function has only two kinds of values so that the Walsh transform has exact expressions and has only two kinds of operations. The Walsh transform can be realized easily and quickly. Automatic face recognition is the main research hotspot of artificial intelligence and mode recognition. Many methods have been used for human face recognition and the method based on the neural network is the main research aspect recently. In order to expand the application of Walsh function, Walsh transform and the fast algorithm was researched by combining the two-dimensional Walsh transform and the neural network for human face recognition. The main content of this paper includes: Firstly, the main theory of Walsh function was reviewed . The one-dimensional and two-dimensional Walsh transform and the fast algorithm was discussed in detail based on the Walsh function definition and creation. Specially, the fast algorithm of Walsh transform based on generalized Gray codes and the fast algorithm of two-dimensional Walsh transform with the inverse Walsh ordering was designed by using the order number method in binary system and expressions decompose method. Secondly, a method based on the neural network was designed for human face recognition by combining the Walsh transform and the characteristic of face recognition, and the experiment result was analyzed. After the proper pre-processing, the recognition rates and effects were improved by combining the two-dimensional Walsh transform and the neural network. Then the two kinds of orderings were compared and it was found that the two-dimensional Walsh transform with the inverse Walsh ordering made energy more focused and made the recognition rates higher than the Walsh-Hadamard transform .