基于盲源分离和噪声抑制的语音信号识别

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。

Speech Signal Recognition Based on Blind Source Separation and Noise Suppression

This paper presents a signal recovery algorithm for the personalized speech command automatic recognition system in the ubiquitous voice environment. This algorithm can separate the mixed speech source from the multiple voice transmitter, and detect the speaker by tracking the high amplitude part of the speech power spectrum and suppressing the noise adaptively. An automatic speech recognition (ASR) is designed and implemented to deal with multiple vocalizations. The experimental results show that the proposed method of source separation and noise suppression does not require any additional computation process and effectively reduces the calculation burden, using the speech database NOIZEUS.