Learning and Application of Feedback Path Database for Feedback Control in Hearing Aids

Abstract

State-of-the-art feedback control systems make use of adaptive filters to model and compensate for the acoustic feedback paths, which vary in different acoustic situations. In a hearing aid application, the acoustic feedback paths are user dependent and vary over time. Although there are repeatable patterns in these variations, it remains to be seen that modern feedback control systems actively learn from and then use these patterns to improve the adaptive filter performance. In this letter, we present a new idea of learning and storing acoustic feedback paths as hearing aid users wear their hearing aids, and then to apply these learned and stored acoustic feedback paths to control and support the adaptive filter updates, thereby to achieve significantly improved feedback control performance.