Identification of the stress and relaxation level in people, based on the study and the advanced processing of physiological signals related to the activity of the autonomic nervous system

Abstract

The objective of this thesis is the development and implementation of intelligent algorithms for the real-time processing of non-invasively acquired physiological signals to automatically predict the continuous level of stress and relaxation in people. Thus, be able to identify the activity associated with the autonomic nervous system, responsible for the alterations caused in the homeostatic balance within the body. This goal resulted in a solution that goes from the analysis and processing of physiological signals to the design of an algorithm for real-time prediction of the level of stress and relaxation, which has subsequently been implemented in a functional low-cost hardware prototype. More precisely, the physiological records used to carry out this development are the electrocardiogram, the galvanic skin response and breathing due to their relation with the activity of the autonomic nervous system and the possibility of being acquired non-invasively.The proposed methodology focuses on four main aspects. The first is the processing of physiological signals in short-term sliding windows, which contributes to improve the techniques used for the extraction of heart period through the design of novel algorithms focused on the robust analysis of theelectrocardiogram and blood pressure signals. In the second, the analysis, normalization and labeling of the extracted physiological parameters is carried out using original and validated methodologies. In the third aspect, the resulting data are subsequently employed for the design and training of intelligent systems through the implementation of supervised and unsupervised learning techniques in order to carry out a robust prediction of the level of stress and relaxation. Among the validated methods fuzzy logic, fuzzy rule-based supervised learning systems and artificial neural networks stand out. Finally, the development is successfully implemented in a portable low-cost hardware solution consisting of a physiological signal acquisition module and a server that processes and transfers the information to the client safely in real-time.