Development of a non-parametric probabilistic model of the human middle ear

Abstract

The middle ear is the part of peripheral auditory system that works to transmit the sound energy from the outer to the inner ear, matching their highly different impedances. Due to its importance for the human hearing, the middle ear has been studied through mathematical models over the past 70 years. The use of deterministic approaches has been the mainstream on the development of these models despite it is well recognized that middle ear has a natural variability with respect to its mechanical and dynamical properties, kwon as random uncertainties. In this work a lumped-element model of the middle ear is used as a baseline deterministic model for the development of a probabilistic model using a non-parametric approach. This probabilistic approach is characterized by adding the uncertainties directly into the global matrices of the modeled system. Furthermore, an optimization process using a single objective function was used for fitting both probabilistic and baseline deterministic models. The probabilistic model developed presents promising results, showing statistical responses in agreement with experimental evidences. In addition, advantages were identified in comparison to a model previously developed with a parametric probabilistic approach.