An automatic acoustic feedback suppressor system for public address systems.

Abstract

This report mainly focuses on the study of acoustic feedback cancellation in public address systems using adaptive filters. The problem of acoustic feedback has existed since the invention of public address systems but nothing was being done about it due its complexity. Sound engineers used to mitigate it using traditional means of setting up microphones and speakers a great deal of distance apart however this did not solve the problem effectively. They later resorted to use of filters for example the notch filters, low pass, high pass and band pass filters. The method used in this project was a cascaded system of adaptive filters. The method included two stage that is to say the noise reduction stage (NR) and the acoustic feedback cancellation stage (AFC). In the NR stage, the multi-wiener filter was used and, in the AFC, an adaptive NLMS-PEM is used. Acoustic feedback cancellation (AFC) is a technique used to reduce the negative effects of acoustic feedback in public address systems. Acoustic feedback occurs when the sound produced by the loudspeaker is picked up by the microphone and amplified, resulting in a loud, howling noise. AFC systems typically use an adaptive filter to estimate the acoustic feedback path and then subtract the estimated feedback signal from the microphone signal. The prediction error method (PEM) is a technique that can be used to reduce the bias in the estimation of the acoustic feedback path. PEM works by first modeling the incoming signal as a white noise sequence. The inverse model of the incoming signal is then estimated and used to pre whiten the inputs of an adaptive filter. This pre-whitening step helps to reduce the bias in the estimation of the feedback path. The normalized least mean squares (NLMS) algorithm is a simple and efficient adaptive filter algorithm that is commonly used in AFC systems. However, the NLMS algorithm can suffer from high bias when the incoming signal is correlated with the feedback signal. The NLMS-PEM algorithm is a hybrid adaptive filter that combines the advantages of PEM and NLMS. The PEM algorithm is used to reduce the bias in the estimation of the feedback path, while the NLMS algorithm is used to achieve fast convergence. Simulation results show that the NLMS-PEM algorithm has good convergence rate and tracking performance. The NLMS-PEM algorithm is a promising new approach for acoustic feedback cancellation in public address systems