一种基于盲源分离的语音保留算法  

A Speech Retention Algorithm Based on Blind Source Separation

作　　者：李民行 庞明睿 刘剑[1] LI Minhang;PANG Mingrui;LIU Jian(College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016

出　　处：《电声技术》2024年第12期95-100,共6页Audio Engineering

基　　金：南京航空航天大学研究生科研与实践创新计划项目(xcxjh20230326)。

摘　　要：佩戴隔音耳罩在隔绝噪声的同时隔绝语音信号,为将环境中的语音信号直接传入耳罩内部,提出一种基于盲源分离的语音保留算法。该算法利用盲源分离的方法,将环境中的混合信号分解为噪声信号和语音信号,之后根据信号短时能量特性,给出语音信号的判别条件,利用该条件实现语音信号的保留。通过仿真实验,设定语音信号判别的阈值条件,最后在多种噪声类型和信噪比环境下进行仿真,成功分离并保留噪声环境下的语音信号,验证了该语音保留算法的有效性。Wearing noise-canceling earmuffs blocks both noise and speech signals.To directly transmit environmental speech signals into the earmuffs,this paper proposes a speech Retention algorithm based on blind source separation.The algorithm uses blind source separation to decompose the mixed environmental signals into noise and speech signals.Then,based on the short-term energy characteristics of the signals,conditions for identifying the speech signal are defined and used to preserve the speech signal.Through simulation experiments,the identification conditions are determined.Finally,simulations are conducted under various noise types and signal-to-noise ratio environments,successfully preserving the speech signal and verifying the effectiveness of the speech-preservation algorithm.

关 键 词：语音保留 盲源分离 信号的短时能量 

分 类 号：TU112.3[建筑科学—建筑理论] TU112.59