期刊名称：International Journal of Advanced Research In Computer Science and Software Engineering
印刷版ISSN：2277-6451
电子版ISSN：2277-128X
出版年度：2013
卷号：3
期号：5
出版社：S.S. Mishra
摘要：Speech enhancement is a topic of interest for many years. In particularly, enhancement of speech signals recorded using signal channel devices such as mobile phones is of prime interest in this paper. It is because for these devices, it is not possible to record noise signals separately, and t he surrounding background noises are picked up by the microphone simultaneously with the speech signal. This may even completely fade-in the speech signal, depending upon the signal-to-noise ratio (SNR). To address this problem, numbers of algorithms and techniques have been developed. The key techniques include: spectral subtraction algorithm, statistical model based algorithm such as Wiener filter and MMSE algorithms, and subspace algorithm. However, as these systems are based on certain assumptions and constraints that are typically dependent on the application and the environment, the existing methods are not able to perform homogenously across all noise types. Therefore, for this purpose, auto-correlation function in time domain and frequency domain is used. The magnitude of auto-correlation coefficients is usually large between 2ms and 12ms, as the human pitch period is typically constrained between these values. However, the same is generally not true for noisy speech signals. The auto-correlation function of a noisy speech signal is usually confined to lower time lag and is very small or zero for higher time lag. Therefore, higher-lag auto-correlation coefficients are relatively robust to additive noise distortion. This paper is focused on enhancing the noisy speech signal from single channel devices by using only the higher-lag auto-correlation coefficients. The efficiency of the algorithm is evaluated in terms of coherence estimate of speech signal.
关键词：Single channel speech enhancement; speech processing; spectral subtraction; auto-correlation and ;Speech signals SN R
