基于时频能量比的入侵事件识别方法

Intrusionevent recognition method based on time-frequency energy ratio

针对挖掘入侵事件与人步行等干扰事件的识别问题,提出一种基于时频能量比的识别方法.利用时域的节律特征以及信号包络的时域冲击特征,剔除如车辆路过、自然环境干扰等事件.留下挖掘和人步行事件.对于挖掘和人步行事件的识别,首先,对事件信号进行时域窗分割;其次,将时域分割后的每个子信号输入到一组窄带滤波器中,并计算每个滤波器输出信号与输入的时域子信号的能量比值,得到信号的时频能量比特征.最后,利用SVM作为分类器,进行分类实验.实验表明,该方法提取的时频特征所包含的冗余特征数据量小,分类所需的时间短,分类识别的准确率约为94%.

In order to identify digging intrusion event and interference events such as people walking,a recognition method based on time-frequency energy ratio is proposed.Using the rhythm characteristics of the time-domain and the time-domain impact characteristics of the signal envelope,events such as vehicle passing and natural environment interference are eliminated,digging and human walking events are remained.For identifying digging and human walking events,first,time-domain window segmentation is performed on the event signal.Secondly,each sub-signal after time domain segmentation is input into a set of narrow-band filters,and the energy ratio of each filter output signal and input are calculated,then get time-frequency energy ratio characteristic of the signal.Finally,the SVM is used as a classifier.The experimental results show that the time-frequency features extracted by the method contain small amount of redundant feature data,short time required for classification,the accuracy of classification recognition is about 94%.