表情分析中嘴形的计算机识别

首先建立六曲线嘴部形态模型,并通过亮度均衡、基于HSV表述的图像二值化、外唇线初定与精化、牙齿包络线确定和内唇线认别等一系列步骤和算法,提出嘴部形态模型中各曲线的提取方法。最后采用4阶多项式对上唇外线外的5条唇形曲线进行拟合,并在Matlab下进行了识别实验。结果显示：提出的嘴形识别方法对不同的嘴形具有较高的鲁棒性,唇形曲线识别结果与人工理解结果基本一致。

Computer Recognition of Mouth Figuration for Facial Expression Analyses

A novel six-curve mouth model is firstly proposed in this paper. And through a set of steps and algorithms such as intensity balance, image thresholding based on HSV representation, exterior lip contours detection and refinement, teeth outline identification and inner lip contours recognition, a new ab- straction approach for all six curves is then discussed in detail. At the end of this paper, curve fitting utili- zing a four order poly for all five curves except the exterior upper lip contours curve is adopted. Experiments are taken under Matlab environment and the results show that the proposed mouth recognition approach is robust to kinds of different mouth shapes. The recognition results of lip curves match human understanding at a high degree.