Robust Fractional-Order-Based Von Mises Subband Adaptive Filtering for Feedback Cancellation in Hearing Aids

Abstract

The issue of acoustic feedback poses a consistent challenge in the context of hearing aids since it imposes restrictions on the attainable amplification levels and has the potential to significantly diminish the perceptual audio quality through the generation of whistling sounds. In recent studies, researchers have employed delay-less multiband structured band-wise acoustic feedback alleviation techniques for hearing aids. However, the resilience of this system in the existence of non-Gaussian noise has yet to be thoroughly addressed. To tackle this matter, a robust and new cost function has been introduced based on the modification of the von Mises distribution with a scaling parameter. In addition, fractional lower-order models provide resilience to heavy-tailed distributions, enhanced precision, heightened versatility and suitability for diverse data formats. In light of this viewpoint, a robust fractional order von Mises-based subband acoustic feedback compensation technique is introduced for hearing aids. The method’s efficacy has been evaluated by computer simulations involving speech and music signals at various signal-to-noise ratio (SNR) levels. The evaluation results show 2–3 dB decrease in misalignment and 2–3 dB enhancement in added stable gain in contrast to prior methodologies. Moreover, perceptual evaluation of speech quality and hearing aid speech quality indexes have been improved by $\approx ~2$ % and $\approx ~10$ % respectively as employed with the proposed technique.