Research And Implementation Of Acoustic Feedback Suppression

In real life,the sound reinforcement system is widely used in large conference rooms,multimedia classrooms and variety shows and so on.Acoustic feedback refers to a phenomenon that out speech return to microphone after refraction of wall in the sound reinforcement system.Acoustic feedback will not only lead to distortion of the output signal,but also cause howling when meeting certain gain and phase conditions,and affect the performance of sound reinforcement system seriously.So the conference,teaching and performance and other activities can not be carried out normally.A lot of times,people have to reduce the gain of sound reinforcement system in order to avoid howling phenomenon,so sound reinforcement system can not achieve the ideal sound effect.Therefore,the research and implementation of acoustic feedback suppression technology,improving the quality and stability of the sound reinforcement system output signal system constantly,has a very important practical significance.This paper is intended to achieve an economical and practical acoustic feedback suppression system,to suppress howling and improve the quality of the output signal and the stability of the sound reinforcement system effectively.For this purpose,the following works are mainly done in this paper:Firstly,the principle of howling occur is analyzed and the basic principles of frequency shifting algorithm,adaptive feedback cancellation algorithm and notch-filter-based howling suppression algorithm are studied.By comparing the three algorithms,notch-filter-based howling suppression algorithm is selected to research in depth.And howling detection,spectrum zoom and notch filter design which are the key steps of notch-filter-based howling suppression algorithm are focused on.Then,howling detection is researched in depth.Two novel howling detection algorithms are proposed after analyzing the change features of power spectrum of the howling component.One howling detection algorithm is based on the feature of interframe spectral flatness deviation.The other is based on the feature of long-term signal variability at specific frequency band.Simulation results show that compared with the traditional howling detection method,the probability of false alarm can be controlled more effectively while keeping a high probability of detection in the proposed algorithms.Lastly,a simulation environment based on personal computer platform is built and a hardware platform based on TMS320VC5509 A and TLV320AIC3104 is designed.The implementation and test of the notch-filter-based howling suppression algorithm is completed in hardware platform.The test results show that howling component can be detected and suppressed accurately.Both stability of sound reinforcement system and output speech quality are assured.The whole performance of the howling suppression is satisfactory.