The research on Face Recognition based on improved BP Neural Network

摘要：

This paper study on Face Recognition based on improved BP Neural Network. In the current study on Face recognition, the original sample data was not preprocessed and the feature was not extracted, so that accurate identification decreased. And the complex structures of BP neural are the shortcomings in the current study. Adaptive Particle Swarm Optimization Algorithm has been put forward. Firstly, the original gray-scale image data has been compressed and the noise has been eliminated by use of multi-resolution analysis of wavelet transform, so the low-frequency wavelet subbands has been got. Secondly, a new adaptive particle swarm optimization (APSO) replaced gradient descent algorithm in the BP learning algorithm to get face detection and achieve a face detection system. Finally, simulation results show that, this identification model is simple, and it improves convergence speed and correct detection rate, it also makes the model significantly improved classification performance. The experiments show that the improvement of algorithms in face detection system has achieved good results.

关键词： Neural network Multi-wavelet transform Face Recognition

作者: Xuezhang Zhao Jianxi Peng Xianghua Yang Yunjiang Xi

作者单位: Foshan Polytechnic College Foshan, Guangdong province, China South China University of Technology Guangzhuo, Guangdong province, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 45-48

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）