Algorithm Research Of Adaptive Howling Suppression And Its DSP’s Implementation

The acoustic feedback problems exist in the theatres, multimedia classrooms, conference rooms and other public addressing systems. It often causes significant performance degradations in sound reinforcement system; in the worst case, the systems become unstable and howling occurs. Therefore, acoustic feedback is a big issue that needs solving in sound reinforcement system.Traditional methods of howling suppression, such as improving rooms’ acoustic property or cascading equalizers in the public addressing system, are inconvenient to operate, enhance the system gain only a little and damage the fidelity of acoustic signal.Phase-Modulation Methods and Gain Reduction Methods are flexible, but difficult to get good balance among real-time processing, system gain and the fidelity of acoustic signal. Moreover, FSM always detect and process howlingafter occurring, which makes the users uncomfortable.Acoustic feedback suppression(AHS) overcomes other methods’ shortcomings, does little harm to speech signals, improves the systems’ gain prominently and is easy to operate.The thesis focuses on AHS. Based on in-depth analysis on the theory of adaptive algorithm, the adaptive howling suppression algorithmsand the decorrelation techniquesare discussed. Firstly, the thesis introduces the basic principle of adaptive filter and LMS/NLMS/VMLMS/VSNLMS algorithms. Then, to avoid a biased and slowly converging feedback estimation, the thesis gives an introduction of four kinds decorrelation techniques, including Noise injection, Inserting delay, Half-wave rectification and Frequency shifting. After that, the thesis studies adaptive linear prediction for real-time application. Although AHS algorithms have become common in public addressing systems, there is no standardized objective procedure available for evaluating them. The thesis discusses objective measures for evaluating AHS algorithms from the system performance, the maximum stable gain and the sound quality.In order to provide a virtual environment for the simulation of AHS, the thesis proposes a MATLAB platformand introduces the simulation of AHS. Based on the evaluation criteria of simulation results, the thesis adopts NLMS and VMLMS adaptive howling suppression algorithm, designing the AHS schemausing the fixed DSP chip(TMS320DM6437) of TI and other relevant peripheral equipment. Not only does the AHS schema implement in a virtual sound reinforcement system, but also in a real system. The tests shows that the AHS schema being implemented on DSP can suppress the howling well in the meantime get better performance in evaluation.