Adaptive hear-through filter design and implementation for earphones

Abstract

Hear-through (HT) systems enhance the perception of ambient sound by generating pseudo-ambient sound, and have become essential for modern smart earphones. However, conventional HT design methods assume that passive isolation of earphones has no effect at low frequencies, resulting in low-frequency amplification. To overcome this limitation and improve HT performance, we propose an efficient adaptive HT system. The system introduces a primary path in the HT controller design that characterizes the impact of passive isolation over the full-frequency. We further develop the HT based on filtered-x normalized least mean square (HT-FxNLMS) algorithm, which dynamically adjusts controller coefficients to adapt to ambient sound variations. Extensive simulations and experiments demonstrate superior performance and robustness compared to existing fixed-coefficient algorithms across different acoustic environments. Furthermore, the proposed HT controller enhances environmental speech signals, enabling natural communication for earphone users in real-world scenarios.