This dissertation proposal focuses on applying HAR to ergonomics. Its main goal is to segment upper body postures using multiple time series of human motion (retrieved by inertial sensors, i.e., acceleration, angular velocity, magnetic field, and determined using signal processing and biomechanical modelling, i.e., orientation and joint angles, methods). The student will develop a supervised machine learning model to classify upper body postures, using multiple time series data, and segment postures in time.
In this work, it is expected that the student develops activity recognition algorithms applied to time series. The outcomes consist of the possibility to classify and segment work-related upper body postures conducted within assembly processes, which is aligned with Industry 4.0 project - Operator (on going) - and HfPT PRR project (on going).
Author: Diogo Sozinho
Type: MSc thesis
Partner: FCUL - Faculdade de Ciências da Universidade de Lisboa