Development and evaluation of energy expenditure estimation and activity classification algorithm

Goal

Develop low power machine learning algorithm for energy expenditure and activity classification. MATLAB high-level algorithm implementation was ported to C language and demonstrated on Android wear watch.

  • Funding source(s): Bosch Sensortec GmbH
  • Total project cost: 69,330 EUR
  • Start and end dates: 09.2015 – 04.2016
  • Project manager: Ivo Fridolin

Implementation of research outcomes

Project was conducted in cooperation with Bosch Sensortec GmBH and TalTech dep. of computer systems and dep. health technologies.

The outcome of the project was an ultra-low power smart hub consisting of programmable 32-bit microcontroller, a state-of-the-art 3-axis accelerometer and a powerful software framework containing pre-installed sensor fusion and other sensor processing software within a small LGA package. Some of its functionalities include step counting, activity recognition, walking, jogging, and other activities.