基于改进Gabor小波变换的人脸情感识别

Face emotion recognition based on improved Gabor wavelet transform

为提高Gabor小波提取图像特征的全面性和准确性,设计一种改进的Gabor小波变换算法进行人脸情感识别。该方法通过像素点的融合,增强了每个像素的局部关联性,采用局部保持主成分分析(LPPCA)算法进行降维处理,提取了更全面的特征信息。在标准表情数据集JAFFE上进行试验,结果显示,在使用相同的分类器的情况下,使用该改进方法提取特征的人脸情感识别准确度达到96.36%,比原始的Gabor小波提取特征的方法提高了1.48%,由此可见,该方法提高了Gabor小波提取图像特征的全面性和准确性,在人脸情感识别方面为行之有效的方法。

In order to improve the comprehensiveness and accuracy of Gabor wavelet extracts image feature,an improved Gabor wavelet transform algorithm is designed for face emotion recognition.This method enhances the local correlation of each pixel through the fusion of pixel points,and adopts the local retention principal component analysis(LPPCA)algorithm for dimensionality reduction processing,which can extract more comprehensive special information.Experiments on standard expression datasets JAFFE show that,with the same classifier,the accuracy of facial emotion recognition using this improved method is 96.36%,which is 1.48%higher than the original Gabor wavelet extraction feature.It can be seen that the method improves the comprehensiveness and accuracy of Gabor wavelet image extraction features and is an effective method for face emotion recognition.