Research Of Adaptive Howling Suppression Algorithm Based On Matlab And Howling Suppression Evaluation

Acoustic feedback exists in all kinds of public addressing systems. Howling, whichresults from the feedback, not only restrains the gain increasing, but also makes thepublic addressing systems cannot work when feedback goes seriously. In a soundreinforcement system, the microphone signal consists of a source signal component anda feedback signal component, the latter denoting the entire signal that is fed back fromthe loudspeaker to the microphone. In the field of electro acoustic systems, speechcommunication, visual net meeting system and active noise control system, acousticfeedback is a factor that cannot be ignored. It affects system performance, such aslimiting gain of amplifier, damaging the sound quality, reducing effect of active controlsystems, destroying the stability of system, etc. Therefore, acoustic feedback is a bigissue that needs solving.Recently, with great developments of digital signal processing technology, a newsolution appears—adaptive howling suppressor. It overcomes shortcomings oftraditional means, and it is easier to operate and improves the gain prominently and doeslittle harm to acoustic signals. However, most of the adaptive howling suppressors indomestic market are imported from oversea and are expensive. So if we developsacoustic feedback suppressor with own intellectual property, it will be greatlycompetitive advantage. This thesis is aimed at solving the howling problem andanalyzed the adaptive howling suppressor. A simulation platform is proposed toresemble the real acoustic environment. Then the LMS/NLMS/VMLMS algorithmsare introduced and applied into the simulation scenario. In order to get betterperformance, a new algorithm is proposed—duplicate LMS algorithm.Owing to there is no standard evaluation for the howling suppression, thecontemporary evaluation rules are studied and the evaluations are used to assess oursimulation results. The tests shows that adaptive suppression algorithms can suppress the howling well in the meantime get better performance in evaluation.