A data analysis framework for motion sickness recognition

Motion sickness arises when the brain receives discordant sensory information from the body, eyes and vestibular system. People who suffer from this disorder experience nausea, fatigue and general discomfort when they travel by plane, car, ship, but also in amusement park rides and in virtual reality environments. The aim of this thesis project is to preventively detect this state of illness when travelling, in order to improve the travel experience, but also to prevent any negative effects on their own health. To achieve this objective, different techniques of data analysis are adopted. These physiological signals are recorded during a driving simulation carried out in a laboratory. Data relating to the dermal conductivity of the subjects examined are processed. In the framework outlined by this thesis work, the first point addressed is that of data pre-processing, in order to accurately extract the features that characterize the behaviour of the selected signal. The extraction of the features of the dermal conductivity signal of each subject is performed on overlapping time windows, in order to monitor in detail the physiological response of the subject to each perceived stimulus. During the simulation the individuals provides information about their state of health. To each window corresponds a label assigned on the basis of this information supplied, allowing the classification process. The label identifies whether the subject is in a state of sickness or not in that time window. The proposed solution compares different classification algorithms aiming to build a machine learning model to recognize and therefore predict the motion sickness. The experimental evaluation shows that it is feasible to develop a promising model in order to achieve the explained purpose.