基于热释电红外信息的人体运动特征提取与识别  

Extraction and Recognition of Human Body Motion Features Based on Pyroelectric Infrared Information

作　　者：万柏坤[1] 冯莉[1] 明东[1] 王璐[1] 邱爽[1] 徐瑞[1] 綦宏志[1] 王威杰[2] 

机构地区：[1]天津大学精密仪器与光电子工程学院,天津300072 [2]邓迪大学运动分析研究所

出　　处：《纳米技术与精密工程》2012年第3期249-256,共8页Nanotechnology and Precision Engineering

基　　金：国家高技术研究发展计划(863计划)资助项目(2007AA04Z236);国家自然科学基金资助项目(30970875;90920015;61172008;81171423);国家自然科学基金委-英国爱丁堡皇家学会联合研究项目(30910494);天津市科技支撑计划重点资助项目(07ZCKFSF01300);国际科技合作专项(08ZCGHHZ00300)

摘　　要：运动人体辐射的红外信号中包含步行姿态及体貌特征等信息,热释电红外(PIR)传感器能够探测人体发出的红外线,在有效范围内实现运动人体的检测.对利用表面安装有菲涅耳透镜的PIR传感器探测人体行走时的红外辐射进行了研究,采用主成分分析(PCA)方法对不同对象、不同速度的频谱数据进行特征提取,并使用支持向量机(SVM)方法进行分类识别,最高识别正确率达到85.38%.研究结果表明,利用PIR传感器获取人体运动信息,并将来自不同对象、不同速度的频谱数据用于分类的方法是可行的.该研究可为安全级别较低场所的生物特征识别提供一种新的低成本的解决方案.The infrared radiation of a walking person contains the information of walking gait and body shape.Pyroelectric infrared（PIR） sensors can detect human body motion within a fairly reasonable distance through its heat emissions.In this study,infrared radiation was detected by a PIR sensor with coded Fresnel lens arrays.Principal component analysis（PCA） method was employed to analyze the data obtained from different objects walking at different speeds for extracting their motion features.Then support vector machine（SVM） was used for classification and the highest probability of correct recognition was 85.38%.The study results show that it is feasible to employ PIR sensors to extract human motion features and to use the spectral data obtained from different objects walking at different speeds for classification.This proposed method also provides a new low-cost solution for monitoring low-security places.

关 键 词：热释电红外传感器 主成分分析 特征提取 支持向量机 模式识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]