基于雷达传感器的连续人体动作识别方法

Continuous Human Motion Recognition Method Based on Radar Sensor

针对基于雷达传感器的离散人体动作识别方法难以得到实际应用的问题,文章提出了一种基于雷达传感器的连续人体动作识别方法。首先对连续动作的雷达回波信号进行预处理得到距离时间域图像。然后通过时频分析得到微多普勒时频谱图像。最后分别采用支持向量机与长短期记忆网络作为分类器进行动作识别。实验结果表明,采用长短期记忆网络作为分类器对人体连续动作进行分类的平均识别准确率达到80.7%,对比SVM作为分类器得到的平均识别准确率高4%。

This paper proposes a radar-based continuous human motion recognition method to solve the problem that it is difficult to realize the practical applications of radar-based discrete human motion recognition methods.Firstly,the echo signal of continuous action is preprocessed to obtain the range-time image.Then the micro-Doppler image is obtained by time-frequency analysis.Finally,Support Vector Machine and Long Short-Term Memory Network were used as classifiers for action recognition.The experimental results show that the average recognition accuracy of human continuous motion classification using LSTM is 80.7%,which is 4%higher than that using SVM.