采用同步挤压小波变换的人体运动姿态分析

摘要

针对人体运动的雷达回波信号特征复杂、不同运动姿态微多普勒频率差异小、难以区分精细特征的问题,提出了一种采用参数可调的同步挤压小波变换(SSTAP)的人体运动姿态分析方法.首先根据实测人体运动数据构建人体运动模型及其雷达回波模型;然后利用SSTAP方法对人体运动模型雷达回波信号进行分解,获得人体各主要部位的时频特征;再通过调整同步挤压小波变换的2个参数获得人体整体回波信号的具有最佳时频分辨率的时频特征,进一步与各部位的人体时频特征比较获得了人体运动姿态的信息.实验结果表明,相比广义S变换(GST)、小波变换(WT)等时频分析方法,基于参数可调的同步挤压小波的人体微多普勒分析结果更加清晰精细,更能反映人体微运动的特征,其微多普勒频率的分辨率比GST、WT分别提高了17％和14％.%A method of human motion posture analysis using synchrosqueezed wavelet transform with adjustable parameters (SSTAP) is proposed to focus on the problems of complicated radar echoes of human motion,small frequency difference of micro-Doppler(MD) to the human motion postures (HMP),and difficulties to distinguish the detailed frequencysignatures of different body movements by conventional methods.An HMP model and a radar echoes model of human micromotions are established from measured data based on motion capture technology.Then,the SSTAP method is used to decompose the radar echoes signals and the time-frequency characteristics of the main parts of bodies are obtained.The time-frequency characteristics of the body's overall echo signal with the best resolution are obtained through adjusting two parameters of the SSTAP.The information of HMP is finally obtained by comparing the time-frequency characteristics of whole human body with those of each part of the body.Experiment results and comparisons with the generalized S transform (GST) and wavelet transform (WT) show that,the human MD signatures from the SSTAP are more clear and easily recognized,and the SSTAP is better in describing the signatures of HMP and increases the recognizable rate by 17％ and 14％,respectively.