Evaluación automática objetiva de la capacidad auditiva utilizando probabilidades posteriores en máquinas de soporte vectorial

Abstract

In this paper, a novel method for automatic hearing assessment using auditory brainstem responses is presented. It is based on a pattern recognition system, which consists on four stages: data preprocessing, feature generation and extraction, classification and decision. The classification between normal and pathological responses is performed using support vector machines, and decision is made by evaluating its probabilistic outputs. The results show the good performance of the system, based on the high rates of sensitivity, specificity and precision obtained with a reduced amount of data. © Springer-Verlag Berlin Heidelberg 2007.