Physiological Signals Analysis Based on Deep Learning Algorithms for Drowsiness Identification

Physiological signals are an invaluable data source which can provide useful insightsin disease detection, rehabilitation, and treatment. In recent years, these signalshave been widely investigated using Machine Learning and Deep Learning Algo-rithms. Among other applications, Sleep Staging and Recognition are tasks that cangreatly benefit from the usage of physiological signals analysis using Deep Learningalgorithms. For this particular set of tasks, signals such as Electromyogram (EMG),Electroencephalogram (EEG), Electrocardiogram (ECG), Electrooculogram (EOG)and Photoplethysmogram (PPG) are typically used. This thesis work focuseson Drowsiness Identification and proposes a Deep Learning approach to processphysiological signals in an efficient, fast and automated way. The core componentof the framework is a Recurrent Neural Network, that is able to quickly processdata with the ability of learning time dependencies in temporal data.