Speech enhancement for mobile communications

Abstract: Three major applications of speech enhancement have been studied: noise reduction for mobile telephone handsfree accessories, noise reduction for handheld mobile telephones, and short delay noise reduction for frame-less telephone systems. During conversation, both hearing and speaking adapt to the background noise in a noisy environment. However, when the conversation takes place over the telephone disturbances are more annoying. The disturbances are a problem since the brain will not get the extra visual and other background information when interpreting the speech. The basis for all the methods presented in this thesis is a novel spectral subtraction algorithm, a speech enhancement method. The novel method improves the conventional spectral subtraction method by introducing causal properties and reducing fluctuations of the gain filter used in spectral subtraction. Further, one extra microphone is included to improve the estimate of the background noise. Frame-less communication systems are also considered . A new spectral subtraction method, which performs the speech enhancement filtering in the time-domain.