基于CSI的手势识别方法研究

针对传统手势识别方法中存在的易受环境光线影响、用户需携带设备、部署代价高等不足,提出基于信道状态信息的手势识别方法(HandFi),以空中手写10个阿拉伯数字的独特书写模式引起CSI波形变换为基础,引入衡量当前室内环境变化的指标,来表征室内人数变化时子载波波动程度的变化,使用巴特沃斯低通滤波和主成分分析去噪,并利用小波...

RESEARCH ON HAND GESTURE RECOGNITION METHOD BASED ON CSI

The previous hand gesture recognition approaches are susceptible to the light condition and have disadvantages in deployment and highcost.A hand gesture recognition method based on channel state information(HandFi)is proposed.The unique CSI waveform caused by unique writing patterns of 10 Arabic numerals serve as feature space.The algorithm was used to indicate subcarrier fluctuation level with the change of the number of people in indoor environment by introducing an indicator,which can measure the changes of current indoor environment.The principle component analysis and butterworth low pass filtering were used for de-noising effectively and the wavelet transform was used to extract time-frequency domain characteristics of hand gesture.A CNN-GRU model was designed to mine the spatial-temporal characteristics of the CSI series,thus,the dynamic gesture recognition was realized.The results in the two experimental environments show that the average recognition accuracy of HandFi is greatly improved.