基于Kinect的实时人脸识别系统

摘要

实现一个基于Kinect体感器的实时人脸识别系统。利用先进的Kinect传感器跟踪视频序列中的人脸特征参数，并可自动分割。通过Gabor滤波器提取特征，运用PCA、LDA特征降维和最近邻准则等实现最终人脸识别和分类。测试结果表明该系统对于视频中多姿态低分辨率的人脸有较好识别效果。这对于实时人脸识别系统的设计，尤其是对现今新兴的基于体感器的人机交互应用，有重要借鉴意义。%A real-time face recognition system based on Kinect body sensor is implemented.We first use the advanced Kinect sensor to track the face feature parameters in video sequence,and they can be automatically segmented.Through Gabor filter to extract features, applying the principal component analysis and linear discriminant analysis to deduct the dimensionalities of features,and using nearest neighbour criterion,the final face recognition and classification is achieved.Testing results demonstrate that the system has preferably good recognition effect on multi-view low-resolution face in video.This provides an important reference meaning for the design of real-time face recognition system,especially for the rising body sensor-based interactive applications.