Personalized Hearing Loss Compensation for the Next Generation Hearables and Hearing Aids

Recent advances in precision hearing diagnostics and machine learning are enabling a new era of personalized audio processing. The identification of cochlear synaptopathy (CS) as an early contributor to sensorineural hearing loss (SNHL) has motivated the development of novel, non-invasive diagnostic tools to complement standard audiometry. At the same time, novel machine-learning approaches have lifted the performance of noise reduction in hearing aids and introduce new avenues for audio processing that can compensate for different aspects of SNHL. However, integrating these insights into effective, individualized strategies for hearing aids and hearables remains challenging. Numerical approaches based on auditory modeling offer promising solutions. Computational auditory models can simulate the effects of SNHL on speech processing, providing insight into how audio processing may compensate for specific pathologies. When combined with modern machine-learning techniques, these models can be optimized so that simulations of normal and impaired auditory processing can guide the training of audio processing algorithms. These advanced nonlinear approaches may provide audio processing that partially restores impaired auditory function without relying on the prior assumptions underlying conventional hearing-aid processing, such as prescribed gain rules, wide dynamic-range compression, or fixed attack and release times. This paper provides an overview of existing and novel hearing loss compensation methods and an example of how precision hearing diagnostics, auditory modeling, and machine learning can be integrated to create personalized audio processing methods. DNN approaches to hearing diagnostics and treatment have the potential to drive the next generation of hearing aids and hearables, enabling targeted intervention even at the earliest stages of hearing damage.