Clasificación de niveles de estrés utilizando señales fisiológicas

Historically, the term stress has been used to describe the different reactions that the human body presents under pressure. These unconcious responses occur in order to presserve the individuals integrity and could appear in human signals such as heart activity, sweating or pupil dilatation. The main goal of this Bachellor Thesis is to provide a system that classifies different levels of stress using physiological signals. For this purpose, an experiment that causes stress has been designed. The experiment consists in four incongruents tests of increasing difficulty performed by 24 participants. Their reaction time and correct answers will be used to approach the performance of the participants to the Yerkes-Dodson curves. The more stress the participant is suffering, the worse is the performance. Two physiological signals have been acquired aiming to classify different stress levels: Electrocardiogram (ECG) and Galvanic Skin Response (GSR). The sensors used to capture these signals were SHIMMER3 ECG UNIT and Q sensor respectively. Heart activity is extracted form ECG signal and electrodermal activity is extracted from GSR signal. The features that characterize the heart activity are related to the heart rate, while the features extracted from the GSR signal are related to the sweating. The classification has been developed applying Logistic Regression (LR) and KNN using leave-one-out cross validation (LOOCV). Two classifications will be proposed: relax and test stress levels and relax, concentration and anxiety levels. After obtaining the results, the stress levels have been succesfully classified using a KNN model with K = 3. The accuracy of the system is 75.38% for two stress levels and 63.08% for three stress levels, achieving the main goals of the thesis. According to the LR model, it can also be concluded that the features that provide more information about stress are heart rate variability and the area under the curve of the GSR signal.Ingeniería de Sistemas Audiovisuale