Multiparametric Study of Physiological Signals for the Recognition of Sleep Disorders

Abstract

This dissertation presents a multiparametric study of physiological signals for the recognition of sleep disorders, which are common health issues with severe impacts on quality of life and public health. This study involves an in-depth analysis of sleep dynamics, classification of its stages, and associated disorders using advanced techniques in biosignal processing and machine learning. The research begins by examining the medical context of sleep, including its definition, different stages, and major related disorders such as insomnia, sleep apnea, and movement disorders during sleep. The study also focuses on the role of polysomnography (PSG) in evaluating these disorders through the analysis of electroencephalography (EEG), electrocardiography (ECG), electrooculography (EOG), and electromyography (EMG) signals. This study adopts a multi-signal approach for accurate sleep stage classification, utilizing advanced machine learning techniques such as the K-Nearest Neighbors (KNN) algorithm and wavelet functions for feature extraction. Classification accuracy is further improved by selecting optimal features using deep learning networks and enhancing model performance through techniques such as Gaussian noise augmentation. The results show that integrating multiple physiological signals and analyzing them at different levels significantly enhances sleep stage classification accuracy and the detection of associated disorders. This approach represents a promising step toward developing intelligent, cost-effective, and highly efficient diagnostic tools, contributing to improved diagnosis and treatment of sleep disorders.