An automatic acoustic feedback suppressor for public address systems
Abstract
Over the past few decades, there has been a great deal of research on the issue of acoustic feedback in public address systems. Acoustic feedback is a common problem in public address systems, in which sound waves from a speaker are picked up by a microphone and amplified, and then sent back to the speaker. This creates a positive feedback loop, in which the sound waves are amplified more and more each time they pass through the system. This feedback loop can result in an audible "squealing" or "howling" sound, as the system becomes overloaded with sound. The most promising technology for acoustic feedback control in public address systems is adaptive feedback cancellation (AFC). AFC systems work by estimating the room impulse response, which is the time it takes for sound waves to travel from the speaker to the microphone. Once the room impulse response has been estimated, the AFC system can subtract an estimate of the feedback from the microphone signal. This effectively eliminates the feedback, preventing the squealing or howling sound from occurring. To improve the performance of AFC systems, it is often necessary to decorrelate the loudspeaker signal and the speech signal. Decorrelation means making the two signals less similar to each other. This can be done by using a variety of techniques, such as the Prediction Error Method (PEM) algorithm. The PEM algorithm works by estimating the correlation between the two signals and then using that information to decorrelate the signals. The PEM algorithm has been shown to be effective in improving the performance of AFC systems. In this project, the PEM algorithm was used to decorrelate the loudspeaker signal and the speech signal in an AFC system. The results of the project showed that the PEM algorithm was able to significantly improve the performance of the AFC system, resulting in a reduction in the amount of feedback.