基于小波熵的低误报率人体热释电红外信号识别

Low false alarm rate human body recognition using pyroelectric infrared signal and wavelet entropy

提出一种基于双密度双树复小波变换小波熵特征的热释电红外(PIR)信号人体识别方法。首先对人体和狗的PIR探测器输出信号进行去噪预处理,然后提取信号的双密度双树复小波变换的小波熵作为特征,最后采用最小二乘支持向量机对特征进行分类。实验结果表明:所提取的特征及分类方法对人体与狗的热释电红外信号的识别率可达93.6%。因此该识别方法能大大降低PIR探测器的误报率,并可进一步提升PIR探测器在安防和智能家居系统中应用。

A method for human body recognition using pyroelectric infrared （PIR） signal based on wavelet entropy （WE） of double-density dual-tree complex wavelet transform （DD-DT CWT） is proposed in the paper. The valid data is obtained by removing noise from original signal and then the wavelet entropy of DD-DT CWT is calculated. Least square support vector machine （LS-SVM） classifier is adopted to classify the feature vectors. Experiment results show that the proposed method has good ability to recognize human body and dog and the recognition rate is up to 93.6%. Therefore, the presented method can highly decrease false alarm rate and improve the application area of PIR detectors in security system and smart home system.