自适应啸叫抑制算法的研究与DSP实现

发布时间：2018-06-21 08:11

本文选题：啸叫抑制 + 自适应算法　；参考：《电子科技大学》2014年硕士论文

【摘要】：声反馈是出现在剧院、多媒体教室、会议室等公共扩声系统中的常见问题,它经常使音频扩声系统的性能发生显著衰退,极端情况下会使得系统变得不稳定,发生啸叫。抑制声反馈是扩声系统亟待解决的问题。从改善房间声学环境入手或在扩声器中加入均衡器的传统方法通常操作不便,并且费用较高。相位调制法和增益降低法是比较灵活的啸叫抑制方法,但是在实时性、扩声增益提高以及音质损失之间很难获得很好的平衡,且多为在啸叫发生后进行检测和处理,影响用户的主观听觉感受。自适应啸叫抑制法克服了相位调制法和增益降低法的缺点,能够实现实时处理,同时大幅提高系统增益,带来较小的声音失真,而且硬件成本较低。本论文以自适应啸叫抑制法为主要研究对象,在深入分析自适应算法理论的基础上,对自适应啸叫抑制算法和去相关技术进行了讨论和研究。论文首先介绍了自适应滤波器基本原理,并重点研究了LMS、NLMS、VMLMS以及VSNLMS算法。随后研究了消除信号相关性的去相关技术,包括噪声注入法、插入延时法、时变处理法和非线性处理法。接着阐述了利用自适应线性预测进行啸叫抑制的原理。由于现阶段对啸叫抑制系统还没有统一的测评标准,本文从系统性能、放大能力、音质失真三方面采用多种评价标准结合的方法对啸叫抑制进行评价。为了便于模拟啸叫发生的声场环境,论文搭建了MATLAB啸叫抑制仿真平台,并在此仿真平台中对基于自适应线性预测的自适应啸叫抑制算法及去相关技术进行仿真和分析实验结果。考虑到计算复杂度、扩声增益和音质失真等因素,论文选择了NLMS算法和VMLMS算法进行DSP实现。硬件平台选择以TI公司的定点数DSP芯片(TMS320DM6437)为核心的EVM板,论文分别验证了DSP算法在模拟声场和真实声场中的啸叫抑制性能。为实现DSP与MATLAB仿真平台的实时连接,本文设计了DSP与MATLAB的通信机制。经过大量仿真测试和实际声场测试,验证了本文的啸叫抑制方案能够对信号实时处理,啸叫抑制效果较好,并且能获得良好的声音质量。
[Abstract]:Acoustic feedback is a common problem in public sound reinforcement systems such as theatres, multimedia classrooms, meeting rooms and so on. It often makes the performance of audio reinforcement system decline significantly, and in extreme cases the system becomes unstable and howls. Acoustic feedback suppression is an urgent problem in sound reinforcement system. The traditional method of improving the acoustic environment of the room or adding equalizer to the loudspeaker is usually inconvenient and expensive. Phase modulation and gain reduction are flexible howling suppression methods, but it is difficult to obtain a good balance between real-time, acoustical gain improvement and sound quality loss, and most of them are detected and processed after howling occurs. Affect the user's subjective sense of hearing. The adaptive howling suppression method overcomes the shortcomings of the phase modulation method and the gain reduction method, and can achieve real-time processing. At the same time, it can greatly improve the system gain, bring less sound distortion and lower hardware cost. In this paper, the adaptive howling suppression method is taken as the main research object. On the basis of in-depth analysis of the adaptive algorithm theory, the adaptive howling suppression algorithm and de-correlation technology are discussed and studied. In this paper, the basic principle of adaptive filter is introduced, and the LMSN LMS VMLMS and VSNLMS algorithm are studied. Then the de-correlation techniques are studied, including noise injection method, insertion delay method, time-varying processing method and nonlinear processing method. Then the principle of roar suppression using adaptive linear prediction is described. Since there is no uniform evaluation standard for howling suppression system at present, this paper uses a variety of evaluation criteria to evaluate howling suppression from three aspects: system performance, amplification ability and sound quality distortion. In order to simulate the sound field environment, a MATLAB simulation platform for howling suppression is built in this paper. The simulation results of adaptive howling suppression algorithm based on adaptive linear prediction and de-correlation technology are simulated and analyzed in this simulation platform. Considering such factors as computational complexity, sound amplification gain and sound quality distortion, NLMS algorithm and VMLMS algorithm are selected for DSP implementation. The hardware platform is based on TI's fixed-point number DSP chip TMS320DM6437. the performance of DSP algorithm in simulated sound field and real sound field is verified in this paper. In order to realize the real-time connection between DSP and MATLAB, the communication mechanism between DSP and MATLAB is designed. Through a large number of simulation tests and actual sound field tests, it is verified that the proposed scheme can process the signal in real time, and the roar suppression effect is better and the sound quality can be obtained.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.3

【参考文献】