Deep Learning Models for Physiological Data Classification of Children During Computerized Auditory Tests

Abstract

The chapter aims to classify the physiological data of hearing impaired (HI) and typically developed (TD) children using machine/deep learning techniques 1) to reveal if the physiological data of the HI and TD are distinguishable, 2) to understand which emotions of HI and TD are recognized, and 3) to investigate the effect of computerization in a subset of audiology perception tests. Physiological signals, which are blood volume pulse (BVP), skin conductance (SC), and skin temperature (ST), are collected using a wearable E4 wristband during computerized and conventional tests. Sixteen HI and 18 TD children participated in this study. An artificial neural network (ANN) and a convolutional neural network (CNN) model are used to classify physiological data. The physiological changes of HI and TD children are distinguishable in computerized tests. TD children's positive (pleasant) and negative (unpleasant) emotions (PN) are distinguishable on both computerized and conventional tests. HI children's neutral and negative (unpleasant) (NU) emotions are distinguishable in the computerized tests.