为了提高特征提取环节对表情识别率的影响,文中采用活动外观模型(AAM)提取整体形变信息,对眉毛及眼睛区域采用Cabor小波变换提取纹理信息,对嘴巴区域采用模板匹配法获取嘴部纹理信息,然后对提取的各个特征采用离散的隐马尔科夫模型得出6种表情概率,在识别阶段根据每个特征对6种表情的贡献权值分别进行特征加权融合,最后选择最大概率的表情作为表情识别结果.通过对10位女性6种表情图像进行训练实验,该方法有着良好的识别率.%In order to improve the impact of the feature extraction on the rate of the face recognition,overall deformation features are extracted by AAM, texture features of the eyebrows and eyes areas are extracted by Gabor wavelet transformation, and template matching is used to extract the texture features of the mouth area. And then discrete HMM is adopted to get the expression probability of each feature. In the stage of recognition, the results are fused by the contribution to the six kinds of expressions of each feature with its weight obtained by contribution analysis algorithm, and then choose the maximal probability as the final result. Through the training and experiment on 10 women of their 6 kinds of expression, the method has good recognition rate.
展开▼
