Detekce pohybových aktivit měřených EKG holterem s akcelerometry

Abstract

This thesis focuses on the detection of movement activities based on data measured by an ECG Holter device equipped with accelerometers. A literature review was carried out, examining technologies for motion detection and methods for classifying movement activities. Models of convolutional and recurrent neural networks were proposed and implemented to recognize walking, running, sitting, and lying down. The results were evaluated on a training dataset and subsequently validated using real-world data provided by the company BTL. Furthermore, a software application was developed for uploading, processing, and visual inspection of the data. During testing, factors affecting the quality of detection were identified and assessed.