Modelagem de Sistemas Audiom\'etricos Usando T\'ecnicas de Computa\c{c}\~ao Flex\'ivel

TL;DR

This paper develops a flexible computational approach using interval data, neural networks, and the MQ estimator to model audiometric systems, effectively managing uncertainties and improving system reliability.

Contribution

It introduces a novel method for modeling audiometric systems with interval parameters, combining neural networks and the MQ estimator to handle data uncertainties.

Findings

Interval parameters enable modeling with uncertain data.

Models validated with two prediction methods.

Interval results encompass most validation data.

Abstract

In systems identification, the studied phenomena are accompanied by uncertainties, whether arising from measurement data or computational calculations. Interval data provides a valuable way to represent available information on complex problems where uncertainty, inaccuracy, or variability must be taken into account. The present work aims to determine interval parameters for a model considering data measurement uncertainties using the MQ estimator and neural networks. The main objective of this work is to apply this technique in audiometric systems, particularly in the automatic detection of auditory responses taking into consideration the flexible computing concepts that offer solutions tolerant to subjectivity, inaccuracy or uncertainty. It was possible to obtain a model with interval parameters that allow an infinite set of parameters to be evaluated as a limited range. The model was validated using two methods, one with prediction of 2 steps ahead, using the MQ estimator and extended to another representation as neural networks 5 steps delay. It has been shown that interval parameters generate interval results that contain most validation data improving system reliability.