基于概率协作表示的多表情序列融合识别

Multi-Expression Sequence Fusion Recognition Based on Probabilistic Cooperative Representation

传统表情识别往往是基于单一图像进行特征提取、训练及识别,缺乏在动态时间上的细微表情变化提取。利用时间前后的人脸表情变化信息,提出了一种基于概率协作表示的多视频序列融合表情识别方法。先采用主动外观模型(AAM)定位出人脸表情的68个特征点,利用提出的融合策略将相邻3帧表情图像的AAM特征进行融合,最后利用概率协作表示的分类优势进行识别。实验结果表明,在CK+表情数据库上,所提出的方法能够把握表情的时间变化信息,相比于近几年的表情识别算法,在识别率上取得了较好的效果。

Traditional facial expression recognition often uses a single image to extract features, train, and recognize;however, subtle changes in dynamic facial expressions are not recognized. This study proposes a multi- expression sequence fusion recognition method based on probabilistic cooperative representation using the changes in facial expression before and after time. First, 68 feature points of facial expression are located using an active appearance model (AAM). Then, the AAM features of three adjacent facial expressions are combined using the the proposed method. Finally, the classification advantages of probabilistic cooperative representation are used for recognition. Experimental results indicate that the proposed method can grasp the temporal change information of expression on the CK + expression database. Moreover, this method can achieve higher recognition rates compared with traditional expression recognition algorithms.