基于MFCC加权动态特征组合的声纹识别技术在地下电缆防护的应用  

Application of Voiceprint Recognition Technology Based on MFCC Weighted Dynamic Feature Combination in Underground Cable Protection

作　　者：李伟 曾繁洋[1] 王博 陈忠斌 LI Wei;ZENG Fanyang;WANG Bo;CHEN Zhongbin(School of Electrical and Automation Engineering,East China Jiaotong University,Nanchang 330013,China;Shanghai Institute of Microsystem and Information Technology,Chinese Academy of Sciences,Shanghai 200050,China;Zhongke Weibo(Suzhou)Intelligent Technology Limited Company,Suzhou 215615,China)

机构地区：[1]华东交通大学电气与自动化工程学院,江西南昌330013 [2]中国科学院上海微系统与信息技术研究所,上海200050 [3]中科伟博(苏州)智能科技有限公司,江苏苏州215615

出　　处：《电力信息与通信技术》2022年第5期16-22,共7页Electric Power Information and Communication Technology

基　　金：国家电网有限公司总部科技项目资助“基于多参量复合感知的特高压输电线路绝缘子串数字化研究与应用”(520950220002)。

摘　　要：地下电缆在使用过程中,容易受到工程车辆作业破坏,针对这种情况,文章提出将声纹识别技术应用在地下电缆防护中,通过收集环境声音,判断电缆周围是否存在工程车辆作业,及时向有关部门发出预警。在声纹识别过程中,提出加权动态梅尔频率倒谱系数(mel frequency cepstrum coefficient,MFCC)作为声纹识别的特征参数,由梅尔倒谱系数及其加权的一阶和二阶差分系数合并而成,不仅能够反映音频信号的静态特性,还在很大程度上逼近音频的动态特性,并且与常用的特征参数相比,计算复杂程度更低。通过实验证明,以加权动态MFCC特征参数作为高斯混合模型–通用背景模型(gaussian mixture modeluniversal background,GMM-UBM)的特征参数的工程车辆识别系统,能够在提高识别率的情况下,大大降低系统的运行时间,以此实现地下电缆防护。In the process of use,the underground cable is easy to be damaged by the construction vehicle operation,in view of this situation,this paper applies the voiceprint recognition technology in the underground cable protection,by judging whether there is an engineering vehicle operation around the cable by collecting environmental sound,and timely warn to the relevant departments.In the process of voiceprint recognition,the weighted dynamic mel frequency cepstrum coefficient(MFCC)is used as the characteristic parameter of voiceprint recognition combined by MFCC and its weighted first and second order difference coefficient.The proposed weighted dynamic MFCC not only can reflect the static characteristics of the audio signal,but also largely approximate the dynamic characteristics of it.Compared with the commonly used feature parameters,its calculation complexity is lower.The experiment has proved that the engineering vehicle identification system with weighted dynamic MFCC feature parameters as the characteristic parameters of Gaussian mixture modeluniversal background(GMM-UBM)can greatly reduce the running time of the system while improving the recognition rate,so as to achieve the role of underground cable protection.

关 键 词：地下电缆防护 声纹识别 MFCC 加权动态MFCC GMM-UBM 

分 类 号：TM75[电气工程—电力系统及自动化]