基于小波和余弦变换的神经网络人脸识别法

Face recognition based on wavelet transform DCT and BP neural network

提出了利用小波变换和余弦变换与 BP 神经网络相结合的人脸识别方法。将人脸图像归一化后进行小波变换,再用余弦变换对低频信号提取特征向量,达到降维和去除干扰的目的,并把特征向量送进 BP 神经网络训练。识别时,对人脸图像进行相同的变换后,送入神经网络进行辨别。实验结果表明,该算法优于传统的人脸识别法。

It presents a face recognition algorithm that is based on wavelet transform, discrete cosine transform and BP neural network. The images normahzed are decomposed using wavelet transform. Then the low frequency coeffcients are transformed into discrete cosine coefficients which are taken into BP neural network as feature vectors to be trained. At the stage of the recognition, the images are decomposed using the same transform, and sent to the neural network which is trained to be recognized. The experiment shows that the face recognition method presented by this paper is better than the traditional one.