Multimodal Workplace Monitoring for Work-Related Fatigue Recognition

Summary

The aim of this thesis is to design and implement a system that deals with the recognition of human behavior on issues of a mental nature. Since the input of technology in human daily life is growing more and more every day and so, new applications and methods of computer science to enter it. Therefore, there is a growing interest in creating applications that will serve as an aid for people experiencing problems with work fatigue, stress and anxiety. In essence, this work is a modern application field of data science, which combines the disciplines of information technology and psychology. In this work, an approach is presented with the aim of automatically recognizing activities that take place, within a work environment, through a series of observations which include actions, behaviors and situations that are a source of fatigue, anxiety and stress. In order to achieve the goal of this work, a combination of information sources is used, which come from the peripheral parts of a computer, as well as information coming from force sensitive resistor sensors. In addition, in order to extract knowledge various machine learning techniques, and more specifically supervised learning are explored. Initially, a set of activities is defined, which take place within a working environment. Based on these activities, measurements are collected, which are used to construct a data set that reflects the activities performed during the working time. Various classifiers are then trained to identify the various activities.