Investigating the impact of using image representations of motion sensor data on human activity recognition.

Motion sensor data consists of time-series signals. With the increasing popularity in the computer vision domain, the idea of representing time series as 2d images is gaining attention and providing promising results. However, there are different methods to encode time series as images, and each may perform differently on different datasets. This project will investigate the impact of using different "encoding time series as image" methods with different parameter settings.

Project co-advisor: Sümeyye Ağaç