Automatic annotation of data collected through Inertial sensors

Abstract

The data annotation process is a time-consuming process related to Human Activity Recognition (HAR). The data annotation process is generally done by a human assigned to watch video recordings of activities executed by people and thereafter labels identified activity or activities for each video frame. This thesis involves presenting a potential solution to this time-consuming process by the use of Bluetooth Low Energy (BLE) beacons and smartwatch within the users' environments instead of video recording and thereafter manuel review. The solution utilizes acquired data from sensors in beacons and smartwatches suchlike heart rate, temperature, magnetometer and accelerometer. Accumulation of data must not interfere with the privacy nor daily life of the user during their activities.