基于PCA与SVM的头部姿态识别及应用

Head Pose Recognition and Application Based on PCA and SVM

摘要:

针对头部姿态识别在复杂背景和变化光照情况下准确率低的问题,提出了一种有效识别图像序列中头部姿态的方法.首先运用Adaboost算法提取出图像序列中不同姿态的人脸图像,通过主成分分析方法(PCA)提取人脸姿态特征;然后使用支持向量机构(SVM)造多分类器对提取的特征分类从而实现头部姿态识别;最后设计了五种不同的头部姿态在变化光照下与智能轮椅进行人机交互实验.实验结果表明该方法实时性高,抗光照变化性能强,识别率高达92.2%.

Abstract:

In order to improve the low rate of head pose recognition under complex background and changeable illumination,in this paper,presented is a method which can efficiently recognize the head pose in image sequence.In this approach,faces with different poses are detected using Adaboost algorithm that ensures extremely rapid face detection,and the face features are extracted using principal component analysis(PCA) which remains the global grayscale features of a frame of image.The support vector machine(SVM),which can achieve a high classification accuracy with a few of labeled training samples,is used to classify the head poses.Lastly,five head poses are adopted to interact with an intelligent wheelchair for disabled people.A set of experiments,having an average recognition rate of 92.2%,demonstrate the proposed method’s real-time and robustness against the variations of illumination.