无线传感器网络中声震信号的特征提取算法  

Feature Extraction Algorithm for Acoustic-seismic Signals in Wireless Sensor Networks

作　　者：曹红兵[1] 魏建明[1] 刘海涛[1] 

机构地区：[1]中国科学院上海微系统与信息技术研究所,上海200050

出　　处：《小型微型计算机系统》2010年第2期211-214,共4页Journal of Chinese Computer Systems

基　　金：上海市科委重点项目(07dz15011)资助

摘　　要：面向无线传感器网络在地面目标声震信号识别方面的应用需求,提出基于局域判别基(Local Discriminant Bases,LDB)算法的特征提取方法.并且,针对现有的基于时频能量图的可分性测度的缺点,提出新的基于概率密度估计的相对微分熵的可分性测度.基于实地采集到的信号的仿真实验表明,该方法在一定程度上提高了目标的正确识别率,降低了特征维数,具有实际的应用价值.Feature extraction algorithm based on local discriminant bases （LDB） is proposed to satisfy the requirement of application on classification of acoustic-seismic signals of ground targets in wireless sensor networks. And a new discriminant measure using relative differential entropy which is based on probability density estimation is proposed to solve the problem of time-frequency energy map. Experiments based on real signals indicate that this method can improve the performance of classification at a certain extent and decrease the dimension of features,so it is practically valuable for application.

关 键 词：无线传感器网络 特征提取 LDB 时频能量图 相对微分熵 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]