Kirubagari Balasubramaniam - Academia.edu (original) (raw)

Uploads

Papers by Kirubagari Balasubramaniam

This paper presents a novel speech more reliably than conventional multi-channel schemes, enhance... more This paper presents a novel speech more reliably than conventional multi-channel schemes, enhancement scheme on the basis of bone-conductive in general the additional gains to be derived from them microphones. High performance in non-stationary noisy have not been commensurate with the greatly increased environments is achieved by combining the following two computations required by these methods to combine techniques: the speech detection based on the information from these different sensors. bone-conductive microphone signal, and non-stationary To overcome this problem, this paper proposes a noise suppression by adaptive spectral subtraction. Unlike new speech enhancement scheme for a non-stationary regular air-conductive microphones, bone-conductive noisy environment. Firstly, the bone-conductive microphones are insensitive to environmental noises and microphone based speech detection is used to distinguish background speech. By employing this advantage, we are between background noise segments and speech segments. able to detect very robustly speech in non-stationary noisy Secondly, the reliable speech detection results help to environments. Furthermore, based on the reliable speech determine when to re-estimate the noise threshold. detection results, the adaptive spectral subtraction can Thirdly, the adaptive spectral subtraction aims at operate effectively even under low SNR conditions. attenuating the non-stationary noise. And the final enhanced speech is produced.

This paper presents a novel speech more reliably than conventional multi-channel schemes, enhance... more This paper presents a novel speech more reliably than conventional multi-channel schemes, enhancement scheme on the basis of bone-conductive in general the additional gains to be derived from them microphones. High performance in non-stationary noisy have not been commensurate with the greatly increased environments is achieved by combining the following two computations required by these methods to combine techniques: the speech detection based on the information from these different sensors. bone-conductive microphone signal, and non-stationary To overcome this problem, this paper proposes a noise suppression by adaptive spectral subtraction. Unlike new speech enhancement scheme for a non-stationary regular air-conductive microphones, bone-conductive noisy environment. Firstly, the bone-conductive microphones are insensitive to environmental noises and microphone based speech detection is used to distinguish background speech. By employing this advantage, we are between background noise segments and speech segments. able to detect very robustly speech in non-stationary noisy Secondly, the reliable speech detection results help to environments. Furthermore, based on the reliable speech determine when to re-estimate the noise threshold. detection results, the adaptive spectral subtraction can Thirdly, the adaptive spectral subtraction aims at operate effectively even under low SNR conditions. attenuating the non-stationary noise. And the final enhanced speech is produced.

Log In