基于小波变换和神经网络集成的人脸表情识别

Facial Expression Recognition Based on Wavelet Transformation and Neural Network Ensemble

提出了一种表情识别的新方法,首先通过基于小波变换的图像分解和K-L变换等处理,来抽取面部表情区域的有效鉴别特征,之后采用神经网络集成技术对六种典型表情进行识别。在CMU表情数据库上的实验表明,该方法达到了很高的识别率,而且对光照变化也有一定的不敏感性。

This paper introduces a novel method for human facial expression recognition. We first acquire the low-dimension features by using wavelet transformation and K-L transformation, then an neural network ensemble was construeted to recognize six basic facial expression. Experiments on CMU database show that it performs better than the single neural network, and does not sensitive to the lighting condition.