中文摘要：

当前多数图像序列的人脸表情识别方法仅提取图像的某一类特征，导致特征参数不能全面地反映脸部情感信息．提出一种基于混合特征和多HMM融合的图像序列表情识别方法．采用Gabor小波变换、二维离散余弦变换分别提取眼睛及眉毛区域、鼻子区域的纹理变化特征，对嘴巴区域则采用主动表观模型提取形状变化特征．对待测图像序列中的每个表情特征区域采用离散隐马尔可夫模型得出6种表情概率；然后根据在训练阶段得到的每个表情特征区域对每种表情的贡献权值进行加权融合，并选择融合后的表情概率最大者作为识别结果．实验结果表明，该方法综合了表情的纹理与形状变化，能够得到很好的识别效果，且处理速度快，适合于实时图像序列的表情识别．

英文摘要：

Most of facial expression recognition methods for image sequences generally extract one kind of features currently, which results in a shortage that the features can not effectively reflect comprehensive facial emotional information. A method of expression recognition based on hybrid features and multiple HMMs fusion for image sequences is presented to address this problem in this paper. Texture features for the eye area are extracted by using Gabor wavelet transformation, texture features for the nose area are extracted by using 2D-DCT, and shape deform features for the mouth area are extracted by using AAM. Discrete HMM is adopted for expression recognition in each expression area of the testing image sequences respectively. The recognition results are fused by means of integrating the probability of each expression in each area with its weight obtained by contribution analysis algorithm, and the final expression is determined as that with the maximal probability. Experimental results show that the method can integrate the texture and shape deform features of expressions and get high recognition rate. The method is highly efficient in its running and is suitable for real time expression recognition.