基于ICA与BP神经网络相结合的人脸识别研究

Face Recognition Based on ICA and BP Neural Network

提出了独立成分分析和BP神经网络相结合的人脸识别方法。首先对人脸图像进行小波分解,提取人脸图像的低频子带图像,然后用独立成分分析方法对低频子带图像进行特征提取,并用改进的BP神经网络对所提取的人脸特征进行分类识别。在AR人脸图像库中进行的实验表明,此方法取得了较高的识别率。

A face recognition method combining independent component analysis with BP neural network is presented. At first, the low frequency sub-band images are extracted from original faee images with wavelet decomposition. Independent component analysis is employed to extract face features. Finally, an improved BP neural network is applied to classification. Experimental results on the AR face database show that the method has good pefformance in terms of the recognition rate.