基于频域逐级回归的声学回声控制

发布时间：2018-01-04 16:07

本文关键词：基于频域逐级回归的声学回声控制　出处：《电子与信息学报》2014年12期 　论文类型：期刊论文

【摘要】：传统声学回声控制算法一般采用基于随机梯度法更新的频域分块自适应滤波(PBFDAF)方法,但在以语音为主要回声信号的室内混响环境中,由于回声路径不稳定,往往收敛速度较慢,难以实现足够的回声抑制。该文提出一种基于频域逐级回归的声学回声控制算法。通过逐级回归分析远端信号和麦克风信号之间的线性关系,可以在保持较小的偏差的同时实现收敛较快的系统估计。同时,由于逐级分析了两通道间的短时相干性,因而该算法无需像常见方法一样,额外进行基于通道间相干函数的残余回声抑制或双讲检测,从而保持系统的紧凑性。若进一步假定近端背景噪声准平稳,则可利用基于近端信号非平稳程度的自适应平滑因子,在实现系统估计快速收敛的同时确保其稳定性。实验表明,该方法在常见的近端环境噪声水平下,在收敛速度和稳态误差上相对传统方法有显著优势,非常适合应用在室内远讲模式下的声学回声控制中。
[Abstract]:The traditional acoustic echo control algorithm generally adopts the frequency domain adaptive filter based on the random gradient method, but in the indoor reverberation environment where the voice is the main echo signal. Because the echo path is unstable, the convergence rate is usually slow. It is difficult to achieve enough echo suppression. An acoustic echo control algorithm based on stepwise regression in frequency domain is proposed in this paper. The linear relationship between remote signal and microphone signal is regressed by stepwise regression. The system estimation with fast convergence can be realized while keeping small deviation. At the same time, because the short-time coherence between the two channels is analyzed step by step, the algorithm does not need to be the same as the common methods. Residual echo suppression or dual-talk detection based on the coherent function between channels is performed to keep the system compact. If the near-end background noise is further assumed to be quasi-stationary. The adaptive smoothing factor based on the non-stationary degree of the near-end signal can be used to realize the fast convergence of the system estimation and to ensure its stability. The convergence rate and steady-state error are superior to traditional methods, and are suitable for acoustic echo control in indoor telecontrol mode.
【作者单位】：中国科学院声学研究所;中科院语言声学与内容理解重点实验室;
【基金】：国家自然科学基金(10925419,90920302,61072124,11074275,11161140319) 中国科学院战略性先导科技专项(XDA06030100)资助课题
【分类号】：TN912.3
【正文快照】： 1引言在远讲或免提模式下的语音通信和人机交互系统中,远端语音信号经扬声器播放和房间传播后,会被本地麦克风接收,从而产生回声,严重影响语国家自然科学基金(10925419,90920302,61072124,11074275,11161140319)和中国科学院战略性先导科技专项(XDA06030100)资助课题音通信系

【参考文献】