节点文献
基于CNN-LSTM神经网络的热释电红外传感器人员识别
Person identification based on CNN-LSTM neural network and PIR sensor
【摘要】 针对热释电红外(PIR)传感器在室内人员识别系统的结构以及识别的准确率问题,设计了一种新型的无线分布式PIR传感器系统,并提出了一种人员识别的新方法。系统采用2只分布在不同高度的PIR传感器,结合对菲涅尔透镜的视场角调制,能够有效探测运动人体的红外信号。通过对2只PIR传感器时域输出信号的采集分析,采用快速傅里叶变换(FFT)算法获取时域信号特征,并将信号特征进行融合。使用深度学习卷积神经网络—长短期记忆(CNN-LSTM)神经网络进行人员的分类识别。实验结果表明:该设计方法在人员的分类识别上实现了99.29%的准确率,在室内人员识别场景中具有良好的应用价值。
【Abstract】 Aiming at the problem of structure of pyroelectric infrared(PIR)sensor in indoor person recognition system and accuracy of recognition, a new type of wireless distributed PIR sensor system is designed and a new method of person recognition is proposed.The system uses two PIR sensors distributed at different heights, combined with the field angle modulation of the Fresnel lens, which can effectively detect infrared signal of moving human body.Firstly, the method collects and analyzes the time-domain output signals of the two PIR sensors.Then, fast Fourier transform(FFT)algorithm is used to obtain the characteristics of the time-domain signals, and fuses the signal characteristics of the two sensors.Finally, the deep learning CNN-LSTM neural network is used for person classification and identification.The experimental results show that the proposed method achieves accuracy of 99.29 % in classification and identification of persons.It has good application value in indoor person identification scenes.
【Key words】 pyroelectric infrared(PIR) sensor; pattern recognition; deep learning; person identification;
- 【文献出处】 传感器与微系统 ,Transducer and Microsystem Technologies , 编辑部邮箱 ,2023年01期
- 【分类号】TP183;TP212
- 【下载频次】156