基于自适应阈值滤波和S-Method的穿墙人体动作识别

Through-Wall Human Motion Classification Based on Adaptive Threshold Filtering and S-Method

穿墙人体动作识别在武装反恐、城市巷战、灾害救援、病人监护等领域具有重要的应用价值。传统的基于短时傅里叶变换(Short-Time Fourier Transform,STFT)的时频分析方法时频分辨率低,不利于后期的分类识别。本文提出了一种基于自适应阈值滤波和S-Method的时频特征增强方法,用于墙后人体动作识别。该方法首先利用自适应阈值滤波消除时频图中的噪声,然后采用S-Method方法聚焦能量,提高时频特征,最后利用K最近邻(KNN)分类器对人体动作进行识别。利用频率步进穿墙雷达获取的实验数据进行方法验证,结果表明:相比于传统的STFT方法,本文所提出的方法对走、跑、坐、跳、招手以及原地踏步等6种典型动作的平均识别准确率更高,可达96.11%。

Through-wall human motion classification plays an important role in many applications,such as antiterrorism,urban warfare,disaster rescue,and patient monitoring.The traditional time-frequency(T-F)analysis method based on short-time Fourier transform(STFT)has low T-F resolution,which is not suitable for classification.In this paper,a T-F feature enhancement method based on adaptive threshold filtering and S-Method is proposed for through-wall human motion classification.It firstly uses an adaptive threshold filtering to eliminate the noise,and then aggregates the energy of the T-F map via the S-Method to enhance the T-F features.Lastly,K-nearest neighbor(KNN)classifier is applied to recognize the human motions.Experiment results based on real data obtained by stepped-frequency through-wall radar demonstrate that,for six common human activities including walking,running,sitting,jumping,waving hand and piaffing,the proposed method can achive an average classification accuracy of96.11%,which is higher than that of conventional STFT method.