Research On Human Identification Method Based On WiFi Sensing Technology

With the development of wireless sensing technology,WiFi sensing technology has become a research hotspot in the field of indoor human identification with its many advantages such as non-contact,passive sensing and easy deployment.Since the WiFi signal propagation is easily interfered by the complex multipath effect in the room,a large amount of data preprocessing work are required in traditional feature extraction method,and it is difficult to extract effective features to improve the human identification accuracy.Therefore,it becomes an urgent problem to effectively extract features from WiFi signals and achieve robust identification.In this paper,the influence of gait on Channel State Information(CSI)in WiFi signals is used to identify.The walking state of the target is detected to separate the stationary data,thereby an effective data range for gait-based human identification is provided.The two dimensions of spatial and time are combined based on the CSI data during walking,and gait features are extracted automatically by deep learning to achieve human identification.The main research contents are as follows:(1)In order to distinguish the state of stationary and walking in the indoor environment,and to determine the data range including effective gait behavior,a CSI-based gait behavior detection algorithm WiTread is proposed in this paper.By analyzing the spatial correlation of CSI data in subcarriers,the input matrix for deep learning is established.The two-dimensional convolution operation is used to extract the local spatial features from adjacent subcarriers and perform gait behavior detection,and the workload of data preprocessing is reduced.The experimental results show that WiTread can effectively distinguish the static and walking data in the CSI data,and determine the data range of gait behavior.The foundation for gait-based huamn identification is provided.(2)In order to solve the problem that the gait feature in WiFi signal is difficult to extract efficiently,a deep learning-based huamn identification algorithm WiID is proposed in this paper.The CSI data when the people walking is utilized in thealgorithm,and pooling technology is adopted to combine with the hidden unit information of all time steps in the recurrent neural network,then timing modeling of spatial features from the time dimension is realized.The gait feature extraction of spatial and temporal dimensions is completed,and an end-to-end human identification model is constructed.The experimental results show that the identification accuracy of WiID is between 95.6% and 98.9% when the number of people is 2~6,and it also has good robustness.In this paper,features in CSI data are extracted automatically by deep learning method,and gait behavior detection and human identification are effectively realized.It not only gets rid of the dependence on manual coding features,but also has a high accuracy rate,providing a practical and effective implementation way for end-to-end contactless human identification.