中文摘要：

低复杂度的振动信号常规峰度检测算法只能检测人员入侵行为,不能检测车辆入侵行为。根据噪声数据与车辆入侵数据特点,结合时间窗和马尔柯夫过程概念,提出了基于改进峰度的振动信号检测分类算法。该算法引入背景噪声的平均能量,利用包含比例因子p和异常突变阈值r的分段函数代替信号能量,可以避免毛刺信号干扰,检测并区分人员入侵和车辆入侵而不需要任何先验条件。试验表明,该算法具有复杂度低,资源要求低,漏警率低的特点。

英文摘要：

A conventional kurtosis algorithm for the vibrating signal is low-complexity.It does not affect vehicle intrusion detecting but only human intrusion detecting.Combined with the timing window and the Markov process,an improved algorithm based on the principle of kurtosis is proposed.The algorithm adopts the average energy of background noise replaces the signal energy with the piecewise function contain of the scale factor p.And the abnormal mutation threshold r can avoid the glitch signal interference and classify human intrusion and vehicle intrusion without the requirement of any prior statistical knowledge of signals.Experimental results show that the novel algorithm is more efficient in the signal detection and classification and has advantages of low complexity,low resource requirement and low false alarm rate.