色噪声环境中TLS-ESPRIT谐波谱重构语音增强研究  

Research of Speech Enhancement of Harmonic Spectrum Reconstruction Using TLS-ESPRIT Method in Color Noise Environment

作　　者：邹大勇[1] 王永灿[2] 赵建军[1] 

机构地区：[1]河南大学计算机与信息工程学院,河南开封475004 [2]河南有线电视网络集团,河南许昌461000

出　　处：《电声技术》2010年第2期59-62,共4页Audio Engineering

摘　　要：为了提高语音在色噪声环境中的信噪比,提出了一种基于总体最小二乘旋转不变子空间技术(TLS-ESPRIT)谐波谱重构语音增强方法。通过对观察数据矩阵进行奇异值分解,有效地将信号及噪声分开。运用TLS-ESPRIT算法对语音谐波信号进行谱估计,重构语音信号,消除了帧与帧之间的噪声残留,得到了在巴克域上与原始语音几乎相同的语音信号。In order to raise the single-to-noise ratio of speech in color noise environment, a methods of speech enhancement of the harmonic spectrum reconstruction based on TLS-ESPRIT（Total Least Square Estimation of Signal Parameters Via Rotational Invariance Techniques） is proposed in this paper. Observational data by singular value decomposition matrix can be effectively separated from the signal and noise. To use the TLS-ESPRIT algorithm estimated spectrum about the harmonic wave of speech, and reconstructing speech single can eliminates the frame and the frame between the residual noise, obtain nearly the same as the primitive speech in Barker domain.

关 键 词：谐波谱重构 总体最小二乘旋转不变子空间技术 语音增强 空间谱估计 

分 类 号：TN912[电子电信—通信与信息系统]