一种舰载低信噪比环境下的音频端点检测算法  

An audio endpoint detection method ina low shipborne SNR environment

作　　者：王中正 王鉴[1,2] 韩焱[1,2] 韩星程 WANG Zhongzheng;WANG Jian;HAN Yan;HAN Xingcheng(Shanxi Key Laboratory of Signal Capturing and Processing,North University of China,Taiyuan 030051,China;School of Information and Communication Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学山西省信息探测与处理重点实验室,太原030051 [2]中北大学信息与通信工程学院,太原030051

出　　处：《兵器装备工程学报》2023年第3期197-203,共7页Journal of Ordnance Equipment Engineering

基　　金：国家自然科学基金青年科学基金项目(62203405);山西省基础研究计划(20210302124545);山西省高等学校科技创目(2020L0301);山西省青年科技研究基金项目(201901D211250)。

摘　　要：针对舰载环境下音频端点检测准确率及鲁棒性较低的问题,提出了一种谱减法和朴素贝叶斯分类器相结合的音频端点检测算法。首先提取纯净音频信号MFCC0与GFCC0构建融合特征,与能熵比特征一同作为朴素贝叶斯分类器的输入进行训练及建模,再利用多窗谱谱减法提升待测含噪信号信噪比,提取信号相关特征,朴素贝叶斯分类器根据待测信号特征判断该信号的类别。仿真实验结果表明,该算法针对舰载低信噪比含噪音频信号与传统方法相比有效降低了虚检和漏检,具有更好的准确性及鲁棒性。Aiming at a low accuracy and robustness of audio endpoint detection in a shipborne environment,this paper proposes an audio endpoint detection method combining spectral subtraction and naive Bayes classifier.Firstly,the pure audio signals MFCC 0 and GFCC 0 are extracted to construct fusion features,and,together with the energy entropy ratio feature,they are used as the input of the naive Bayes classifier for training and modeling.Then,the multi-window spectral subtraction is used to improve the SNR of the signals with noise to be measured.The signal-related features are extracted,and the naive Bayes classifier determines the type of the signal according to the characteristics of the signals to be tested.The simulation results show that,compared with the traditional method,the algorithm effectively reduces false detection and missed detection for low SNR frequency signals with noise on shipboard,which has better accuracy and robustness.

关 键 词：音频端点检测 多窗谱谱减法 Mel频率倒谱系数(MFCC) Gammatone频率倒谱系数(GFCC) 朴素贝叶斯 

分 类 号：TJ83[兵器科学与技术—武器系统与运用工程] TN912.16[电子电信—通信与信息系统]