融合全局与局部特征的UWB雷达人体动作识别算法

为了消除雷达信号中杂波和噪声对人体动作识别的干扰，提高小样本数据下动作识别的精度，在去除杂波及噪声干扰的基础上，提出一种融合全局与局部特征的超宽带（ultra-wideband，UWB）雷达人体动作识别算法。用动目标指示（moving target indication，MTI）结合自适应中值滤波对雷达原始回波信号进行预处理，再对人体动作的雷达二维特征图像利用主成分分析（principal component analysis，PCA）提取主要分量作为全局特征表征，并用二维离散小波变换（2D discrete wavelet transform，2D-DWT）结合奇异值分解（singular value decomposition，SVD）获取特征图像在不同方向与尺度划分下动作的局部特征表征，并将全局与局部特征进行串联融合；根据融合特征，在网格搜索算法（grid search，GS）优化的支持向量机（support vector machines，SVM）模型中实现人体动作的识别分类。实验结果表明，该算法能有效获取雷达信号中的人体动作信息，平均识别准确率为95.63%，具有良好的识别性能。

Human action recognition algorithm incorporating global and local features by UWB radar

In order to eliminate the interference of clutter and noise in radar signal on human action recognition and improve the accuracy of action recognition under small sample data, on the basis of removing the clutter and noise interference, this paper proposes an ultra-wideband (UWB) radar human action recognition algorithm based on global and local features. Firstly, the original radar echo signal is preprocessed by moving target indication (MTI) combined with adaptive median filtering. Secondly, the principal component analysis (PCA) is used to extract the main components of the two-dimensional radar feature image of human action as the global feature representation, and the 2D discrete wavelet transform (2D-DWT) combined with singular value decomposition (SVD) is used to obtain the local feature representation of the feature image under different directions and scales, and then the global and local features are fused in tandem. Finally, human action recognition is realized in the support vector machines (SVM) model optimized by grid search (GS). Experimental results prove that the algorithm can effectively obtain human motion information from radar signals, with a recognition accuracy of 95.63% and good recognition performance.