【作者】 张强；

【导师】 黄耀倞；

【作者基本信息】 大连海事大学 ， 信息与通信工程， 2010， 硕士

【摘要】 噪声的存在给语音通信带来干扰,致使通信效果变差。尤其是在强噪声背景环境下,语音信号甚至完全被背景噪声所淹没,几乎无法识别。通信过程无法正常的实现,给生产、生活、军事行动带来严重影响,给个人、单位乃至国家造成巨大的经济损失。因此,在强噪声背景下,如何有效地抑制和消除干扰噪声成为人们研究的热门课题。本文是基于宜昌船舶交通管理系统工程枝城大桥站和巴东站项目中VHF通信语音记录系统的语音处理方法的研究。针对时变强噪声背景环境下语音通信中的噪声消除问题,就现有的去噪方法进行对比分析,最终选定自适应噪声对消,并就自适应噪声对消的可行性进行分析论证。之后,对滤波器结构做了对比分析,对最小均方误差(LMS)算法和递归最小二乘(RLS)算法进行性能比较,最终以有限冲击响应(FIR)横式滤波器结构和递归最小二乘(RLS)算法构建系统模型,确定自适应噪声对消方案。本文以自适应滤波算法的研究为主要内容。在对标准的递归最小二乘(RLS)算法深入分析的基础上,针对其自身固有缺点,提出改进方案。之后,引入三种已经成功应用在自适应均衡和自适应估计的改进算法,并进行原理分析和性能验证。然后,将标准递归最小二乘(RLS)算法和三种改进算法应用在自适应噪声对消系统中。在MATLAB仿真平台下,四种算法分别对语音信号进行去噪仿真实验,并就实验结果进行分析总结。实验结果表明,Kwang-Seop Eom等人提出的基于卡尔曼滤波的具有快速跟踪能力和强抗噪声能力的RLS算法(简称为ISPRLS算法)不仅误差小,而且抗噪声能力强,在强噪声背景下,能取得优异的去噪效果,具有很重要的理论意义和应用价值。更多还原

【Abstract】 Noise interference with voice communications, resulting in the deterioration of voice quality. Especially under the high noise condition, voice signals was even completely overwhelmed by background noise and almost impossible to identify. Voice communications can not be achieved. It has a significant impact on industrial production, daily life and military operations, also makes huge economic losses to the individuals, factories and even the nation. Therefore, under the high noise condition, how can we suppress and eliminate the noise effectively has become a hot research issue.This paper is based on the study of voice processing in VHF communications system,which is one part of the project of the vessel traffic management systems of between Zhicheng Bridge Station of Yichang and Badong station The purpose of this paper is to denoise in voice communications under high noise condition.After the analysis and comparison of existing denoiseing methods, feasibility of adaptive noise cancellation was selected. Then, the finite impulse response (FIR) horizontal filter structure and recursive least squares (RLS) algorithm were chosen to construct the system model after the analysis and comparison of the structure and algorithm.Adaptive filtering algorithm was the main contents of this paper. Based on in-depth analysis of the standard recursive least squares (RLS) algorithm, the improved scheme was proposed for its inherent weaknesses.Three improved algorithms were introduced which has been successfully used in adaptive equalization and adaptive estimation. Then, the standard recursive least squares (RLS) algorithm and the improved algorithms were applied to adaptive noise cancellation system.The above four algorithms denoised the voice signal by MATLAB simulations, which was interferenced by noise. And the experimental results were analyzed and summarized.A new least-squares algorithm based on the Kalman filter was proposed by Kwang-Seop Eom and the other people. The algorithm(ISPRLS algorithm) not only has a fast tracking capability, but also was immunised against measurement noise. Under the high noise condition, it also can achieve excellent denoising effect.The effectiveness of the algorithm were confirmed by computer simulations.更多还原