基于联合模型的人脸识别定位算法仿真研究

研究人脸定位和识别精度问题。由于人脸在拍摄过程中可能出现的形变,位置变化,光照变换等因素影响,造成人脸模糊不清,为了提高人脸识别定位的精确度,提出了一种新的ASM和AAM联合模型迭代的人脸定位和识别算法。首先利用ASM提取人脸轮廓上关键点特征,并对人脸定位。在初始定位的基础上,利用AAM对人脸进行投影,产生训练集合中没有出现的合成人脸数据。以上两步交替进行,产生足够的、稳定的人脸形变图像。识别过程中,将变换矩阵与原始合成数据进行比对。仿真结果显示,改进的方法能稳定地提取人脸轮廓,并准确定位,具有很高的识别效率。

Facial Expression Recognition Based on Refined Unified Model

The accuracy of face detection and recognition problems were studied.The distortion of human face may appear in the film deformation,position change,light change and other factors,which leads to obscure the face.In order to improve the positioning accuracy of face recognition,a new iterative algorithm of human face location and identification was proposed based on ASM and AAM.Firstly,ASM extracted key facial features of the contour as the face location.On the basis of the initial positioning,the AAM was used to project face to produce the training set which did not appear in the synthesis of face data.The two steps were carried out alternately to generate sufficient,stable person face shape variable images.In the recognition process,testing images were also similar to the transformation,and the transformation matrix was compared with the original synthesis data to achieve the purpose of face recognition.Simulation results show that the improved method can extract the face contour and stability of its position.