人脸局部遮挡表情特征快速识别方法仿真

Simulation of Fast Recognition Method for Facial Expression of Facial Partial Occlusion

针对传统的遮挡表情特征识别方法存在的识别速度慢,精度低,误差率大等问题,提出一种基于K-Means算法的局部遮挡表情特征的快速识别方法,采用Gabor滤波器提取人脸的面部表情特征,将滤波器的尺度与滤波器方向与模一起带入提取局部遮挡表情图像多尺度和多方向特征,根据两个融合准则将局部表情图像相同尺度不同方向的特征融合在一起,将融合结果作为LBP算子来描述人脸图像的纹理特征,利用K-Means算法将LBP编码映射至最近的纹理特征聚类中心,构建查找表,快速统计出人脸表情图像的LBP直方图特征,并将其作为局部遮挡表情的鉴别特征,用于局部遮挡表情特征分类识别。实验结果证明,上述方法可以快速、精确地识别出局部遮挡表情特征。

Traditional method has slow recognition speed, low precision and large error rate in facial expression recognition. In this paper, we focus on a method to fast recognize for expression features with partial occlusion based on K-Means algorithm. At first, we used Gabor filter to extract the facial expression features. Then, we combined the scale and direction of filter with the mode to extract the multi-scale features and multi-directional features of facial expression image with partial occlusion. According to two fusion rules, we fused the features with the same scale and different directions and used fusion result as LBP operator to describe the texture feature of face image. After that, we used K-Means algorithm to map LBP codes to the nearest texture feature clustering center. Finally, we built a lookup table to quickly calculate LBP histogram feature of facial expression image, which was regarded as the distinguishing feature of facial expression with partial occlusion for the classification and recognition. Simulation results prove that the proposed method can quickly and accurately recognize facial expression features with partial occlusion.