一种基于Kinect的二维三维混合面部表情识别方法

A 2D and 3D hybrid facial expression recognition method using Kinect

目前面部表情识别的研究和应用大多基于二维图像。文章提出一种基于Kinect设备的二维三维混合面部表情识别方法。通过结合Kinect设备捕捉到的面部特征点位置及动画单元的特征,来确定由面部表情产生的三维网格的形变。应用基于支持向量机的分类方法及基于表情分析和最大置信度的融合算法对表情进行识别。实验表明,该方法比基于二维图像的识别方法性能更好。

Most of the research and application on facial expression recognition are based on 2D images. This paper presents a 2D and 3D hybrid facial expression recognition method using Kinect, which determines the deformation of facial expression 3D mesh according to the feature point positions and features of the animation units captured by Kinect. The facial expression classification is then realized by using support vector machine(SVM) and fusion algorithm based on maximum confidence and emotional profiles. The experimental results indicate that this method has better performance than those recognition methods purely based on 2D images.