一种基于连续波雷达的手势识别方法  

A Hand Gesture Recognition Method Based on Continuous Wave Radar

作　　者：孙延鹏 艾俊 屈乐乐 SUN Yanpeng;AI Jun;QU Lele(College of Electronics Information Engineering,Shenyang Aerospace University,Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学电子信息工程学院,沈阳110136

出　　处：《电讯技术》2021年第7期815-820,共6页Telecommunication Engineering

基　　金：国家自然科学基金资助项目(61671310);辽宁省兴辽人才计划基金项目(XLYC1907134);航空科学基金项目(2019ZC054004);辽宁省百千万人才工程基金项目。

摘　　要：为了解决依靠光学传感器进行手势识别对外部环境依赖较大的问题,提出了一种基于连续波(Continuous Wave,CW)雷达的手势识别方法,并建立了4种手势动作的回波数据库。首先,对CW雷达回波进行短时傅里叶变换(Short-Time Fourier Transform,STFT)获取手势动作的时频谱;然后,通过设立阈值将时频谱中的背景杂波去除;接下来,对处理后的时频谱提取方向梯度直方图(Histogram of Oriented Gradient,HOG)特征;最后,采用支持向量机(Support Vector Machine,SVM)作为分类器,以HOG特征作为输入进行手势识别。实验结果表明,所提方法在普通室内环境下的识别精度超过95%,能够对典型的手势动作进行有效识别。To solve the problem that external environment affects recognition accurary,a new hand gesture recognition method based on continuous wave(CW)radar is investigated,and a dataset of four typical hand gestures echo wave is generated.Firstly,the short-time Fourier transform(STFT)of CW radar echoes from hand gestures is applied to produce the time-frequency spectrograms of hand gestures.Next,the background clutter in the time-frequency spectrograms is removed by setting up thresholds.Then,the histogram of oriented gradient(HOG)feature is extracted from the processed time-frequency spectrograms.Finally,the support vector machine(SVM)is used as a classifier and the HOG features are used as input for gesture recognition.The experimental results show that the recognition accuracy of the proposed method reaches more than 95%in a normal indoor environment,and it can effectively recognize typical hand gestures.

关 键 词：手势识别 连续波雷达 短时傅里叶变换 方向梯度直方图 支持向量机 

分 类 号：TN957[电子电信—信号与信息处理]