基于BP_RRA神经网络的人脸识别
首发时间:2011-04-27
摘要:针对人脸识别问题,以BP算法和RRA理论为基础,采用BP_RRA作为人脸识别分类器,并结合小波变换,提出一种基于小波变换和BP_RRA神经网络的人脸识别算法。首先通过小波变换对人脸图像特征进行降维,然后把分解低频系数作为人脸特征带入BP神经网络进行训练,并用进化算法RRA对BP网络进行优化处理,使其能够快速收敛、避免陷入局部极值。实验结果表明本算法的识别率较高,能够针对不同类别进行有效的训练,对面部表情变化、一般光照及姿态变化具有良好的鲁棒性。
For information in English, please click here
Based on the RRA_BP neural network face recognition
Abstract:For face recognition, base on BP algorithm and the RRA theory, classifier using BP_RRA as face recognition, and combined with wavelet transform, proposed based on wavelet transform and the RRA_ BP neural network face recognition algorithm. First by wavelet transform of face image features dimension reduction, then the decomposition of low-frequency coefficients as facial features into BP neural network was trained, using evolutionary RRA algorithm of BP network optimized, So it can rapid convergence, avoid getting into a local extremum. Experimental results show that the algorithm recognition rate is higher, according to different category can effectively training, changes in facial expressions, face recognition rate significantly increased.
Keywords: face recognition wavelet decomposition BP RRA
基金:
论文图表:
引用
No.4421250583545130****
同行评议
共计0人参与
勘误表
基于BP_RRA神经网络的人脸识别
评论
全部评论0/1000