摘要
为充分提取人脸表情特征并满足在线实时性需求,提出了基于局部投影韦伯梯度编码(Weber Gradient Coding,WGC)描述的人脸表情识别方法。该方法首先计算表情图像当前像素点与周围位置上的数值差同当前像素点的差异构成WGC特征的差动激励,其次通过空间金字塔划分形成表情图像空间子区域,再次分别在水平、垂直和倾斜正负45°上对WGC特征向量进行投影,得到人脸表情局部投影WGC特征,最后通过余弦距离度量表情相似度,并分别在公共人脸表情库JAFFE和CK表情库中进行实验,平均识别率及平均特征提取时间分别为85.6%、12.30 ms和90.96%、28.21 ms。结果表明:该方法较好描述了包含空间分布信息和纹理信息的人脸表情且具有较低的时间复杂度,能满足在线情感识别需求。
In order to fully extract the facial expression features and meet the online real-time requirements,the facial expression recognition method based on local projection Weber Gradient Coding(WGC)description is proposed.The method first calculates the difference between the current pixel and the value difference between the current pixel and the surrounding position of the expression image to form the differential excitation of WGC feature.Secondly,the spatial sub-regions of the expression image are formed by dividing the spatial pyramid.Then,WGC feature vectors are projected on the horizontal,vertical and inclined±45°respectively to obtain the local projection WGC features of facial expressions.Finally,the cosine distance is used to measure the facial similarity,and the experiments are carried out on the public facial expression database JAFFE and CK expression database respectively.The average recognition rate and average feature extraction time are 85.6%,12.30 ms and 90.96%,28.21 ms respectively.The results show that this method can better describe the facial expression containing spatial distribution information and texture information,and has low time complexity,which can meet the need of online emotion recognition.
作者
齐梅
王军丽
刘则芬
QI Mei;WANG Junli;LIU Zefen(Anhui Open University,Hefei 230022,China)
出处
《安徽开放大学学报》
2022年第2期87-92,共6页
Journal of Anhui Open University
基金
安徽省高校自然科学研究重点项目(项目编号:KJ2021A1255)
安徽开放大学青年项目(项目编号:QN202111)。
关键词
局部投影
WGC特征
人脸表情
在线情感识别
local projection
Weber Gradient Coding(WGC)
facial expression
online emotion recognition
