摘要
提出了一种新的噪声谱估计算法,从采集到的含噪声语音信号中直接估计出噪声谱的变化,并且利用一个初始噪声序列进行噪声谱的更新。实验表明,该算法能准确地估计出噪声谱,相比传统的噪声谱估计器,其准确性和实时性显著提高了,也增强了可靠性,特别是在有强背景噪声(输入信噪比很低)和慢变化噪声的情况下性能得到了明显的提高。
In this paper a new method is proposed to estimate the noise in noisy speech. This algorithm can estimate the change of noise spectral from noisy speech directly. By utilizing an initialization noise sequence, it can update the estimated noise spectral instantly. It proves that the algorithm can exactly estimate the noise spectral. Compared with traditional methods, it greatly improves the veracity and real time capability, enhances the reliability of the algorithm, especially under the condition of strong background noise （very low S/N ratio） and slowly changing noise.
出处
《重庆邮电学院学报（自然科学版）》
2006年第1期1-3,共3页
Journal of Chongqing University of Posts and Telecommunications(Natural Sciences Edition)
关键词
噪声估计器
战场背景噪声
噪声级
语音增强
noise estimator
background noise on battle
noise level
speech enhancement
