| With the advent of the era of artificial intelligence,various applications such as smart homes,smart wearable devices,and virtual reality have begun to emerge.Human-computer interaction acts as a bridge between people and devices in the above-mentioned fields.In the development of human-computer interaction,gesture interaction has attracted much attention as the most expressive way of human interaction.Traditional gesture recognition methods are based on data gloves and vision.However,gesture recognition methods based on data gloves need to force users to wear data gloves integrated with various sensors,which brings additional burdens to users.The vision-based gesture recognition method is easily disturbed by natural lighting conditions,and there is also the problem of privacy leakage.The gesture recognition method based on microwave radar not only does not require users to wear data gloves and is not affected by natural light conditions,but also can work around the clock and avoid privacy leakage.Taking into account the above problems,this article has carried out the research of gesture recognition technology based on microwave radar,mainly including continuous wave radar and frequency modulated continuous wave radar gesture recognition technology research.The main tasks of this paper are:(1)For gesture recognition based on CW radar,this article first analyzes and explores the feature extraction of the CW radar echo signal when the target gesture is moving.Based on the principle of Doppler frequency shift,the classic time-frequency is selected.The analysis method uses the short-time Fourier transform algorithm to obtain the characteristic time-frequency diagram of the target gesture movement in the CW radar echo signal,which is reflected as the change of the target gesture speed over time.When analyzing the feature time-frequency graph,based on the discovery of the similar features of the two types of gestures,a topological support vector machine(SVM)model was further proposed to classify the target gestures.(2)This paper also carried out the research of gesture recognition technology based on FMCW radar.This article first uses the Intermediate Frequency(IF)signal of FMCW radar to simulate and calculate the distance and speed information of target gestures.Further,a two-dimensional fast Fourier transform(2D Fast Fouruer Transform,2D-FFT)algorithm is used to obtain a multi-frame range-Doppler map(RD-Map)representing the motion trajectory of the target gesture.Then use the Multiple Signal Classification algorithm to solve the angle information.According to the coherence of multi-frame RD-Map in time,a shallow convolutional neural network model based on time series is proposed to classify target gestures.(3)Set up a gesture data collection system to collect and mark gesture data.An end-to-end gesture recognition system is implemented to test the recognition accuracy of the gesture recognition system in real time.In the experimental scenario,the influence of different condition factors on the accuracy of gesture recognition based on different radar sensors is explored.Finally,the average accuracy of gesture recognition based on CW radar was 93.4%,and the average accuracy of gesture recognition based on FMCW radar was95.5%. |
