Abstract: This paper introduces an algorithm of facial affective features extraction. It preprocesses a still image with facial affective information, extracts affective feature vectors of the expression sub-regions with Gabor wavelet transformation and calculates fractal box dimension and difference fractal dimensions of a facial expression image. These vectors and dimensions are seen as the extracted features. Experiment shows that different affective features are extracted, and the result is better when different subjects display seven basic affectivity, so affective features can be extracted effectively based on Gabor wavelet transformation and fractal dimension.

Key words: pattern recognition, affective feature extraction, Gabor wavelet transformation, fractal dimension

摘要： 提出一种基于Gabor小波变换与分形维的人脸情感特征提取算法，对包含情感信息的静态灰度图像进行预处理，对表情子区域实行Gabor小波变换，提取情感特征矢量，对人脸兴趣区图像求盒维数和差分分形维数，将经过Gabor小波变换所得的特征矢量和分形维数作为所提取的特征。分析比较了不同测试者7种基本情感的识别效果，实验表明该方法能有效提取与情感变化有关的特征。

关键词: 模式识别, 情感特征提取, Gabor小波变换, 分形维

CLC Number: 

• TP391.4