Research And Application Of Gesture Recognition Technology Based On CSI

With the development of wireless technology,behavior recognition based on wireless channel state information has become a hot research topic.By analyzing the change pattern of CSI,gesture recognition,human body activity recognition and positioning can be performed,and it has the advantages of non-contact sensing,low cost,and non-invasiveness.In the Internet of Things era,gesture recognition is a key technology to facilitate human-computer interaction in many smart home applications.This paper conducts a more in-depth study on CSI-based gesture recognition and detection technology from the aspects of CSI dataset construction and deep recognition network construction.The main work of this paper is as follows:1.In terms of data collection and data set construction,Nexus 5 wireless terminal was used to collect and construct two CSI gesture data sets containing 256 sub-carriers,much higher than the 90 sub-carriers collected by Intel 5300 network card,which were used for identification and detection respectively Task.It makes up for the lack of available detection data in existing public CSI datasets.Aiming at the problem that it is difficult to give the start time and end time of gestures in the detection task,an effective event marking method is given.2.A gesture recognition and detection algorithm based on WavegramSpectrogram CNN structure based on CSI data is proposed and implemented.Experimental results show that the proposed twin-tower structure better achieves the representation ability of gesture features.3.A gesture recognition and detection algorithm based on Multi-scale Convolution Augmented Conformer structure is proposed and implemented.Experimental results on public datasets and self-built datasets verify the effectiveness of the proposed algorithm.