Using Computational Optimization for Audio Filter Design to Improve Frequency Selectivity in the Passive Cochlea

Abstract

Hearing loss affects a significant portion of the global population. It is widely accepted that the effective management of hearing loss early in life helps to avoid development delay, and treating hearing loss later on has a significantly positive effect on quality of life. Conventional hearing aids function by exclusively amplifying sound to compensate for a patient’s decreased hearing threshold. However, they do not compensate for the diminished frequency bandwidth that comes along with sensioneural hearing loss. A collection of hydrodynamically-coupled, damped, driven harmonic oscillators are used to simulate the basilar membrane displacement response to an audio signal as a surrogate for directly modeling audio perception. This model, coupled with a high-dimensional, global, gradient-free optimization technique is used to design filtered audio signals which improve the frequency bandwidth in the hearing-damaged cochlea. By substituting an original simple sinusoidal tone with a complex of frequencies that interfere on the cochlea, we show that it is possible to improve frequency selectivity in the damaged cochlea using only an audio filtering technique. This type of process may, in the future, be incorporated with conventional hearing aid technology to further improve the audibility and intelligibility of audio for those with mild hearing loss.