MACHINE-LEARNING CLASSIFICATION OF DAILY LIVING ACTIVITIES IN PARKINSON'S DISEASE

The thesis work was carried out at the Polytechnic University of Madrid as part of a larger project called NETremor. This project involves the development of a methodology for the daily monitoring of tremor in Parkinson's (PD) and essential tremor (ET) patients, using a smartwatch. Classifying daily living activities (ADL) performed by PD patients and secondly to assess during which activities tremor occurs most are important milestone in evaluating the effectiveness of drug therapy. In this thesis work, the main goal was to build up an algorithm based on artificial intelligence models, starting from inertial movement unit (IMU) signals that has been provided by a previous study to classify ADLs as efficiently as possible, and then to validate it by recording new data using a IMU sensor embedded in a smartwatch [1]. The thesis consists of seven chapters: the second provides an overview of the current state of the art in activity recognition based on inertial sensors and the use of artificial intelligence-based methods. The third chapter presents the experimental protocol to collect data, the rationale behind feature selection, and is illustrated the first classification attempt and its results. in the fourth chapter the final classification model is explained. The fifth chapter consists of the validation of the model using data collected in laboratory and hospital and the analysis of the results. Finally, the sixth chapter presents the conclusion and the future development.