A Neural-Network Framework for the Design of Individualised Hearing-Loss Compensation

Sound processing in the human auditory system is complex and highly non-linear, whereas hearing aids (HAs) still rely on simplified descriptions of or loss to restore hearing. Even though standard HA amplification strategies succeed restoring audibility faint sounds, they fall short providing targeted treatments for sensorineural deficits adverse listening conditions. These shortcomings current devices demonstrate need advanced hearing-loss compensation that can effectively leverage non-linear character system. Here, we propose a differentiable deep-neural-network (DNN) framework be used train DNN-based models based biophysical auditory-processing differences between normal-hearing hearing-impaired systems. We investigate different functions accurately compensate impairments include outer-hair-cell (OHC) cochlear synaptopathy (CS), evaluate benefits our trained speech quiet noise. Our results show enhancement was possible all considered cases, with OHC proving easier than CS. Several objective metrics were estimate expected intelligibility after processing, these simulations hold promise yielding improved understanding speech-in-noise listeners who use processing. Since tuned profiles individual listeners, enter an era where truly individualised hearing-restoration developed tested experimentally.