Automatic Detection and Classification of Hearing Loss Conditions Using an Artificial Neural Network Approach

Abstract

The auditory dysfunction is one of the most frequent disabilities, this condition can be diagnosed with an electroencephalogram modality called auditory evoked potentials (AEP). In this paper, we present a machine learning implementation to automatically detect and classify hearing loss conditions based on features extracted from synthetically generated brainstem auditory evoked potentials, a necessity given the scarcity of full-fledged datasets. The approach is based on a multi-player perceptron, which has demonstrated to be a useful and powerful tool in this domain. Preliminary results show very encouraging results, with accuracy results above 90% for a variety of hearing loss conditions; this system is to be deployed as hardware implementation for creating an affordable and portable medical device, as reported in previous work.