基于VMD与改进DAE人车地震动包络信号识别算法

摘　要解决对野外环境中低信噪比的人车地震动信号进行分类时传统模式识别方法应用不便，以及识别率较低的问题，通过基于包络检波、变分模态分解（VMD）和改进的深度自编码器（DAE）的特征提取算法研究了针对该类信号的处理方法和特征提取方法。首先对目标的地震动信号进行希尔伯特变换，获取信号的平滑包络线，然后对包络线进行变分模态分解，并用相关系数对分解得到的IMF信号进行筛选，并将相关度较高的分量加权合成为高信噪比的中间信号，再对其使用改进的深度自编码器中进行特征提取。最后使用泛化性能好的随机森林算法对信号进行分类，从而实现对人车目标的识别和分类。结果表明：该算法对两类目标综合识别正确率较其他传统算法有较大提高。可见该算法针对该类目标有应用价值。

Recognition algorithm of man vehicle ground motion envelope signal based on VMD and improved DAE

In order to solve the problems of inconvenient application of traditional pattern recognition methods and low recognition rate when classifying human and vehicle ground motion signals with low signal-to-noise ratio in the field environment, the processing method and feature extraction method for this kind of signals are studied through the feature extraction algorithm based on envelope detection, variational mode decomposition (VMD) and improved depth self encoder (DAE). Firstly, the target ground motion signal is transformed by Hilbert transform to obtain the smooth envelope of the signal, then the envelope is decomposed by variational mode decomposition, the decomposed IMF signal is screened by correlation coefficient, and the components with high correlation are weighted and synthesized into an intermediate signal with high signal-to-noise ratio, and then its features are extracted from the improved depth self encoder. Finally, the random forest algorithm with good generalization performance is used to classify the signals, so as to realize the recognition and classification of human and vehicle targets. The results show that the accuracy of this algorithm is higher than other traditional algorithms. It can be seen that the algorithm has application value for this kind of target.