融合人脸多特征信息的表情识别系统

基于对处于不同表情中人脸特征差异的分析，发现用同种方法提取面部各部分特征无法达到信息利用度的最大化，会产生有用信息丢失或者冗余计算，降低了算法的识别准确率和运行速度。针对面部表情改变时，变化最大的3个部分——嘴、额头和眉毛在形状、纹理和距离上的差异，提出用模板匹配法提取嘴部特征，用边缘检测法提取额头特征，用外轮廓检测法提取眉毛特征，并综合这三者的输出得到最终面部表情识别结果的多特征提取识别系统。实验结果验证了该方法的稳定性与有效性，该算法无论在识别准确率还是在整体运行速度上都达到了较高的水平。

Facial Expression Recognition System Based on Multiple Feature Integration

Based on the analysis of the characteristics of different expressions in the human face, it was found that using the same method to extract the facial features cannot maximize the availability of information. The lost of useful information and the redundant computing may reduce the accuracy and the performance the algorithm. Aiming at the three specific parts on the face-mouth, forehead and eyebrows-which makes the largest changes in shape, texture and distance when facial expression changes take place, a new multi-feature extraction method is developed. To extract the features, it applies template matching on the mouth, edge detection on the forehead and contour detection on the eyebrows. The integration of the three feature extraction outputs becomes the final result of the system. Experiments validated the algorithm with stability, effectiveness, a high recognition accuracy and fast running speed.