Applying Machine Learning to Sensor Data from Physical Activities

In recent years fitness trackers and other wearables had their public break trough. Both competitive and recreational athletes are currently using wearable devices or use their smartphone to monitor their training activities. These devices are usually equipped with various sensors, including a GPS tracker, a heart rate sensor and accelerometers. Current popular mobile apps and wearables primarily report on directly measurable fitness metrics, such as speed, heart rate and calories spent. In contrast, variables related to fitness, fatigue and injury risk, which can not easily be determined from these data are largely ignored. However, information about for instance injury-risk metrics could enable runners to make smarter training decisions. Currently, machine learning methods are used both in academia and industry to develop models for predicting these variables that are hard to measure. In this project we focus on advancing state-of-the-art in applying machine learning on this data from physical activities and look for new applications.