基于呼吸心跳时序混叠信号的毫米波雷达身份识别

Millimeter wave radar identity recognition based on aliased respiratory and heartbeat time sequence signals

利用毫米波雷达检测人体生命体征,提出一种基于呼吸和心跳时序混叠信号的身份识别方法,构建由卷积神经网络(convolutional neural network,CNN)和长短期记忆网络(long short-term memory,LSTM)串联混合的分类网络用于分类识别.首先使用毫米波雷达分别对受测者进行回波数据采集;其次通过频谱分析和带通滤波器预处理出呼吸和心跳信息,最终组成三种数据集,每种数据集分别拥有3200个样本序列;再将数据集送入分类网络进行身份信息识别.实验结果表明,所提出的分类模型对呼吸和心跳时序混叠信号样本较其他两种单独体征时序信号样本相比能够更高效的分类识别出受测者,分类结果平均识别准确率达到98%以上.

The paper,with millimeter-wave radar to detect human vital signs,proposes an identity recognition method based on the temporal mixing of respiration and heartbeat signals,and builds a classification network consisting of convolutional neural network(CNN)and long short-term memory network(LSTM)in tandem for classification recognition.Firstly,the echo data are collected from the subjects be means of millimeter wave radar;secondly,the respiration and heartbeat information are pre-processed by spectral analysis and band-pass filter to form three data sets,each with 3200 sample sequences;and then the data sets are fed into the classification network for identity recognition.The experimental results show that the proposed classification model can classify and identify the subjects more efficiently than the other two separate samples of respiratory and heartbeat temporal overlap signals,and the average recognition accuracy of the classification results is over 98%.