Implementação e aprimoramento do método cepstral para cancelamento de realimentação acústica

Abstract

Acoustic feedback is a problem that affects the performance of sound systems, causing deterioration of sound quality and limitation of achievable amplification. Among all the approaches to minimize this problem, stands out the acoustic feedback cancellation methods (AFC), which uses an adaptive filter to identify the feedback path and remove its influence from the system. However, as shown in this work, the application of traditional adaptive filtering algorithms in AFC methods presents a bias in the estimation of the impulse response of the feedback path and a slow convergence speed, affecting the performance of the system and necessitating the development of specific methods. Currently, the most promising solution is the acoustic feedback cancellation method based on the cepstral analysis of the error signal (AFC-CE). The present work explores this method through its implementation and evaluation in simulation environment, achieving an increase of maximum stable gain of up to 30 dB in the amplification system, results similar to those found in the literature. It then presents the development of techniques that make the method more robust to variations in the acoustic environment. Finally, it improves the performance of the AFC-CE method through an algorithm that transforms the forgetting factor into a time variant, increasing in general the stability margin of the sound systems.