Using an Artificial Neural Network for Predicting Reaction Time Based on Physiological Data
Summary
In this thesis findings are described in relation to the accuracy of predicting an individual’s reaction times based on his physiological variables with the use of an Artificial Neural Network. It also includes a reflection on the feasibility of this method. In the course of this exploratory research 5 participants were tested on a custom reaction time task as their Electrocardiographs (ECG), Galvanic Skin Response, and the activity of the masseter (jaw) and corrugator (eyebrows) muscles in the face were monitored. Using cross-validation, it was found that mean error rates are around the 0.3 mark, and minimum errors are often near-zero. This is a very promising result and merits further exploration, as such a system of reaction time prediction can function as a safe- guard for people working in highly-demanding situations. This technique might be able to timely advise them about a lower expected performance from themselves before they decide to take up any dangerous or demanding tasks. This can be relevant for professions like astronaut, firefighter or surgeon. Throughout the paper, astronauts are the focus group, as this research was inspired upon the premise to enrich the MECA tool suite that aims to assist astronauts on long-duration missions.