基于SVM与C4.5混合分类器的人脸表情识别

Facial expression recognition based on C4.5 and SVM hybrid classifier

针对人脸表情识别特点,首先使用几何特征与基于灰度共生矩阵纹理特征的低维混合特征提取方法来提取易分类表情图片的特征向量,然后使用C4.5决策树分类器识别出脸部变化较为明显的表情,再使用SVM支持向量机对较难分类表情图片进行分类。实验结果表明,通过改变决策树算法剪枝中错误样本比率的方法可以获得最佳预测精度,识别率达到90%。

The geometric feature extraction is combined with the low dimensional hybrid feature extraction which is based on gray co-occurrence matrix to extract the features of easy-classified expression images,and then C4.5decision tree classifier is used to identify the obvious changes of expressions.At last,SVM is applied to identify the difficulty-classified expressions.The experimental results show that the optimal prediction accuracy is obtained with error-sample rate in the decision tree pruning algorithm,and the identification rate is up to 90%.