Facial Expression Recognition of a Speaker Using Thermal Image Processing and Reject Criteria in Feature Vector Space
Facial Expression Recognition of a Speaker Using Thermal Image Processing and Reject Criteria in Feature Vector Space
复制标题
使用热图像处理和特征向量空间中的拒绝标准进行说话者的面部表情识别
DOI：
10.1007/s10015-013-0136-7
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
and M. Tabuse
中科院分区：
文献类型：
--
作者：
Y. Nakanishi;Y. Yoshitomi;T. Asada;and M. Tabuse
在我们以前开发的说话人面部表情识别方法中，在图像处理中特征向量空间中特征向量的位置是不完美的。导致面部表情错误识别的缺陷往往远离特征向量所属类别的重心。在目前的研究中，忽略的特征向量产生的缺陷，在特征向量空间中使用拒绝标准的方法被应用到面部表情识别。使用所提出的方法，面部表情的两个主题是可区分的三个面部表情的86.8%的准确率为“快乐”，“中性”，和“其他”，当他们表现出的五个故意的面部表情之一，“生气”，“快乐”，“中性”，“悲伤”，和“惊讶”，而这些表情是可区分的78.0%的准确率由传统的方法。此外，该方法有效地判断训练数据是否可接受的面部表情识别的时刻。
In our previously developed method for the facial expression recognition of a speaker, the positions of feature vectors in the feature vector space in image processing were generated with imperfections. The imperfections, which caused misrecognition of the facial expression, tended to be far from the center of gravity of the class to which the feature vectors belonged. In the present study, to omit the feature vectors generated with imperfections, a method using reject criteria in the feature vector space was applied to facial expression recognition. Using the proposed method, the facial expressions of two subjects were discriminable with 86.8 % accuracy for the three facial expressions of “happy”, “neutral”, and “others” when they exhibited one of the five intentional facial expressions of “angry”, “happy”, “neutral”, “sad”, and “surprised”, whereas these expressions were discriminable with 78.0 % accuracy by the conventional method. Moreover, the proposed method effectively judged whether the training data were acceptable for facial expression recognition at the moment.