摘要
宽带语音增强在改善语音质量方面有着广泛的应用。传统的减谱技术虽然去噪明显,但在听觉上留下了不舒适的"音乐噪声",为了在主客观上得到更好的折衷,引入了子空间技术,将带噪信号分解为不相关的信号空间和噪声空间,并结合人耳掩蔽效应,在主客观上得到了较好的折衷。
Wideband speech enhancement is widely used to improve speech quality. Although traditional subtract spectrum method can reduce noise apparently, it also will leave the uncomfortable musical noise. In order to get a better trade off, the subspace technique is introduced, which separates the noisy signal into irrelated signal space and noise space, and then combines with human-being masking effect.
出处
《电声技术》
2009年第5期55-57,共3页
Audio Engineering
关键词
语音增强
子空间
人耳掩蔽
特征值分解
speech enhancement
subspaee
human-being masking effect
eigenvalue decomposition
