一种基于CSI相位差的手势识别方法

A Gesture Recognition Method Based on CSI Phase Difference

随着人机交互(HCI)技术的发展,人体行为的感知和分析成为实现更高层次人机交互的重要一环,尤其是实现与位置无关的手势识别。针对目前基于信道状态信息(channel state information,CSI)的手势识别方法局限于中心链路的问题,提出基于CSI相位差的手势识别方法,充分利用多径效应和空间多样性,识别中心链路和非中心链路上的微弱手势信号。由于CSI相位受时钟不同步和硬件缺陷的影响无法反映环境的变化,采用线性变换算法对其进行校准,并借助MIMO(multiple-input multiple-output)技术获得相位差。在此基础上,利用Hampel滤波器和Savitzky-Golay滤波器滤除异常点和环境噪声。由于频率多样性,根据平均绝对偏差(MAD)选择变化最大的子载波,之后利用动态时间规整(DTW)算法对手势进行分类。实验结果表明,该方法能有效识别中心链路和非中心链路上的微弱手势信号,实现了与位置无关的识别;在中心链路和非中心链路上分别以90%和86.5%的准确率识别6种手势,所需样本量小,识别时间短。

With the development of human-computer interaction(HCI)technology,the perception and analysis of human behavior has become an important part of achieving higher-level human-computer interaction,especially the realization of position-independent gesture recognition.Aiming at the problem that the current gesture recognition method based on channel state information(CSI)is limited to the central link,we propose a gesture recognition method based on CSI phase difference,which makes full use of multipath effects and spatial diversity to identify the weak gesture signal on a central link and non-central link.Since the raw CSI phase cannot reflect the environment change due to the clock is out-of-synchronization and hardware defects,it is calibrated by a linear transform algorithm and the phase difference is obtained by means of MIMO technique.Based on this,the Hampel filter and the Savitzky-Golay filter are used to filter out abnormal points and environmental noise.Due to the frequency diversity,the subcarrier that has maximum change is selected according to the mean absolute deviation(MAD),and then the gestures are classified by the dynamic time warping(DTW)algorithm.The experiment shows that the proposed method can effectively identify weak gesture signals on the central link and non-central link,and achieve position-independent recognition.It can recognize six gestures with 90%and 86.5%accuracy on the central link and non-central link respectively,with small required sample size and short recognition time.