摘要
谱减法是最常用的一种语音增强技术,其特点是计算复杂度低、实时性强、易于实现。谱减法的主要目的是去除语音信号中的噪声干扰,提高语音信号质量。本文在研究基于改进谱减法的基础之上,提出了利用带噪语音的高频区估计噪声谱以及由短时过零率和短时能量组合而成的加权函数去除背景噪声及音乐噪声的语音增强方法。实验表明,这种时频结合的语音增强方法对背景噪声下的语音质量的增强效果明显。
The spectral subtraction method is the most suitable technique for the speech enhancement. It is of utmost its low computation complexity, real-time capability, and easy implementation. The idea of spectral subtraction is to obtain an estimate of the speech spectral level by subtracting the noise estimation from noisy speech. In this case, the phase of the speech is not processed. The main objective of speech enhancement is to reduce noise and to improve one or more perceptual aspects of speech, such as the speech quality.
In this paper, a new speech enhancement method is presented, which uses spectral and time domain processing and achieves a trade-off between effective and low computational load for real-time operations. In the frequency domain, high frequency components of a noisy speech spectrum are averaged and the spectrum of noise is estimated. In the time domain, a weighting function is used to reduce a residual 'musical noise'. Experiment results show the proposed method reduces greatly musical noise, and improves efficiently speech quality.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2003年第2期247-251,共5页
Pattern Recognition and Artificial Intelligence
关键词
语音增强
谱减法
加权函数
语音信号处理
噪声谱
Speech Enhancement, Noise Spectrum, Spectral Subtraction, Weighting Function
