基于改进DTW算法的窨井盖盗损检测方法  

Detection Method of Manhole Cover Theft Based on Improved DTW Algorithm

作　　者：李林丰 薛波 LI Linfeng;XUE Bo(School of mechanical engineering,Jiangsu University of Technology,Changzhou 213001,China)

机构地区：[1]江苏理工学院机械工程学院,江苏常州213001

出　　处：《电声技术》2022年第9期111-116,共6页Audio Engineering

基　　金：国家自然科学基金项目(62003151);江苏省基础研究计划项目(BK20191035)。

摘　　要：针对城市窨井盖被盗或被损坏的现象,提出一种基于改进动态时间规整(Dynamic Time Warping,DTW)算法的窨井盖盗损检测方法。根据窨井盖在正常振动与非正常振动下产生的声纹差异,采用梅尔倒谱系数(Mel-frequency Cepstral Coefficients,MFCC)并整合差分倒谱系数作为特征参数,结合改进的DTW算法进行声纹识别。通过调整动态时间规整函数,将传统DTW算法搜索区域约束为较小面积的菱形,使其达到减小存储空间和缩短识别时间的目的。仿真实验结果表明,铁锤敲击井盖声、车辆碾压井盖声等七种声音类型的平均识别率为81.4%,平均识别速率提高了29.37%。Aiming at the phenomenon that the manhole cover is stolen or damaged in cities, this paper proposes a method for detecting the theft of manhole cover based on improved Dynamic Time Warping(DTW) algorithm. According to the difference of voiceprints between normal vibration and abnormal vibration of the manhole cover(Mel frequency Cepstral Coefficients,MFCC) and differential cepstrum coefficients are used as the characteristic parameters, combined with the improved DTW algorithm to identify voiceprints. By adjusting the dynamic time warping function, the search area of the traditional DTW algorithm is constrained to a diamond with a smaller area, so as to reduce the storage space and shorten the recognition time, and improve the recognition speed. The simulation results show that the average recognition rate of seven kinds of sound types is 81.4%, and the average recognition rate is increased by 29.37%.

关 键 词：差分倒谱参数 动态时间规整(DTW) 声纹识别 梅尔倒谱系数(MFCC) 

分 类 号：TN912.3[电子电信—通信与信息系统]