基于GD-Kmeans和菲涅尔理论的WiFi手势识别方法

针对手势动作幅度较小难以被WiFi所感知到问题，利用菲涅尔衍射理论对最佳动作捕获位置进行推理以增强感知。针对在实际应用过程中需要判断手势何时发生的问题，提出基于高斯分布-Kmeans聚类的GD-Kmeans手势定位算法。在采集到包含手势的信道状态信息（CSI）数据后，使用低通滤波和DWT滤波进行数据降噪，通过定位算法对手势进行定位切出，最终基于动态时间规整（DTW）进行模板匹配实现对五种手势的判别，其准确率达到了93%。

WiFi Gesture Recognition Method Based on GD-Kmeans and Fresnel Theory

In view of the fact that the gesture motion is small, it is difficult to be perceived by WiFi. The Fresnel diffraction theory is used to reason the optimal motion capture position to enhance the perception. Aiming at the problem of judging when the gesture occurs in the actual application process, a gesture localization algorithm based on Gaussian Distribution-Kmeans（GD-Kmeans） clustering is proposed. After collecting the Channel State Information（CSI） data containing the gesture, the data are denoised by low-pass filter and Discrete Wavelet Transform（DWT） filter, and the gesture is positioned and cut by the localization algorithm, and finally the template matching is implemented based on the Dynamic Time Warping（DTW） to recognition the five gestures. The accuracy rate reached 93%.