子空间分解方法在语音增强系统中的应用

介绍了基于子空间分解的单通道语音增强的方法,并利用有声/无声检测的判决算法提高了对噪声估计的准确度,通过KL展开将含噪语音分解为信号子空间和噪声子空间,并在信号子空间中对原始语音信号进行了最优的估计,有效地滤除了数字语音系统中的宽带噪声,并在一定程度上抑制了“音乐噪声”的影响,获得了较好的语音增强效果。

Application of subspace decomposition approach in speech enhancement system

A subspace decomposition approach is presented for suppression of wideband noise in speech en hancement.A method of Speech/Nonspeech Decision in Low SNR Environment has been used to accurately esti mate the noise characteristics.By applying the Karhunen-Loeve Expansion,we decompose the noisy speech into signal subspace and noise subspace.Enhancement is performed by re-moving the noise subspace and estimating the clean signal from the remaining signal subspace.These tests show that the proposed method not only effectively suppresses the wideband noise but also re-moves the musical noise,even if no prior information is given about the noise characteristics or noise level.[