基于盲源分离和噪声抑制的语音信号识别

Speech Signal Recognition Based on Blind Source Separation and Noise Suppression

为了更准确地在噪声环境中对不同语音信号进行识别,提出了一种用于普适语音环境下的自优化语音活动检测(VAD)算法,该算法运用个性化语音命令自动识别系统的语音信号,并能够有效地从多个发声者的混合语音中分离出个体发声者的声音,通过跟踪语音功率谱的较高幅度部分和自适应地抑制噪声来检测发声者的语音信号;设计并实现了一种处理多个发声者任务的自动语音识别(ASR),免去了对干净的语音变化进行先验估计,直接利用噪声本身产生语音/非语音判决的阈值以完成自优化过程;使用语音数据库NOIZEUS进行了评价测试,实验结果表明,所提出的盲源分离和噪声抑制方法不需要任何额外的计算过程,有效地减少了计算负担。

In order to more accurately identify different speech signals in a noisy environment,a self-optimizing speech activity detection(VAD)algorithm for universal speech environment is proposed,which uses personalized speech commands to automatically identify the speech signals of the system.And can effectively separate the sound of individual vocalists from the mixed speech of multiple utterers,and detect the utterer's speech signal by tracking the higher amplitude part of the speech power spectrum and adaptively suppressing the noise.Designed and implemented an automatic speech recognition(ASR)that handles multiple vocalist tasks,eliminating the need for a priori estimate of clean speech changes,and directly using the noise itself to generate speech/non-voice decision thresholds to complete the self-optimization process.The speech database NOIZEUS was used to evaluate the tests.The experimental results show that the proposed blind source separation and noise suppression methods do not require any additional calculation process and effectively reduce the computational burden.