Research On Key Technologies Of Wireless Gesture Recognition Based On Model Method

With the development of smart home,the market has put forward new demands on human-computer interaction methods.The compatibility of WiFi devices with home scenario creates favorable conditions for the development of WiFi based gesture recognition.Although the perception systems that use multiple pairs of transceiver devices to obtain multidimensional information achieves high recognition accuracy,it does not meet the requirements of practical home scenarios in terms of the number and deployment of devices,which is not only inconvenient to deploy but also increases costs.Therefore,this work aims to achieve positionindependent gesture recognition with only one pair of transceiver devices and maintain high recognition accuracy.However,in the scenario of using only one pair of transceiver devices,the gesture information and spatial resolution are very limited,which cause two main problems.Firstly,the gesture information is more susceptible to noise interference and other factors.Secondly,the complete twodimensional trajectory of the gesture cannot be obtained.Only the projection on the direction of the transceiver can be obtained,so the gesture features are easily confused.(1)To address the first problem,we design a gesture detection algorithm based on SPD(Sample Point Distance,SPD)to intercept the stage of gesture more accurately.This algorithm applies the principle of CSI ratio model to find the difference in the trend of gesture in the complex plane when it’s moving or stationary.Its ability to reduce the introduction of static environmental noise.Additionally,a subcarrier selection algorithm based on distinction and fluency is proposed to select subcarriers with more gesture information and less noise interference to increase the proportion of effective gesture information.(2)To solve the second problem,we exploit the different trends of gestures in approaching or moving away from the transceiver pair.We split the dynamic phase variations of different gestures into a series of segments code based on the actions(traverse,approach,and away).Different gestures have specific trends and therefore will have unique gesture codes,thus converting confusing gesture features into distinguishable gesture codes.Finally,a complete model-based position-independent WiFi gesture recognition system is constructed,combining the corresponding gesture detection,subcarrier selection,and gesture coding algorithms.The experimental results demonstrate that the complete gesture recognition system has better performance for different positions,scenarios,and volunteers,achieving a maximum recognition accuracy of 94.67%for six gestures in single position and an average of 89.17%for five positions.