基于字典学习的77GHz雷达人体动作识别

基于视觉的人体动作识别方法对光线和视距环境较高,并且存在侵犯隐私的问题,在应用中有局限性。为了解决这个问题,提出一种基于毫米波雷达和字典学习的人体动作识别方法。首先对人体动作的雷达回波信号进行时频分析得到时频图,再使用两种特征提取方法对时频图进行降维描述,将两种降维后的数据融合,通过LC-KSVD字典学习算法同时学习多特征字典和一个线性分类器,最后根据稀疏系数和线性分类器来识别动作。在此基础上,设计77 GHz毫米波雷达动作识别实验系统,结果表明:算法在10种人体动作数据集上达到了97.7%的识别准确率,可见所提方法实现了对人体动作的准确识别。

Human Motion Recognition by 77GHz Radar Based on Dictionary Learning

The vision-based human motion recognition method has high requirements of light intensity and stadia, and there is a problem of privacy violations, there are limitations in application. In order to solve this problem, a human motion recognition method based on millimeter wave radar and dictionary learning is proposed. Firstly, The time-frequency spectrogram was obtained by time-frequency analysis of the radar echo signals of the human motion. Then, the two feature extraction methods were used to describe the time-frequency spectrogram, and fused the two kinds of features. Simultaneously learned a multi-feature dictionary and a linear classifier by LC-KSVD algorithm. Finally realizes motion recognition base on sparse coefficients and the linear classifiers. On this basis, the 77GHz millimeter wave radar motion recognition experimental system is designed, and the results show that the algorithm achieves 97.7% recognition accuracy on 10 kinds of human motion data sets. It is concluded that the proposed method achieves accurate recognition of human motion.