STA, 6 November 2018 - Researchers at the Jožef Stefan Institute have swept the field in a global machine learning competition aimed at leveraging the movement sensor capabilities of smart phones.
Participating in the Sussex-Huawei locomotion and transportation challenge, 11 researchers of the institute's Department for Intelligent Systems, separated into two teams, won first and second prize.
Using a combination of deep and classical machine learning models, the two teams were the only among 35 participating groups to exceed 90% accuracy with algorithms helping to identify whether the person carrying a smart phone is sitting still, walking, running, cycling or taking the bus, car, train or subway.
The winners were declared at the high-profile Ubicomp conference, which was organised in Singapore this year.
The achievement was also presented today by the heads of the two Slovenian teams; intelligence system department boss Matjaž Gams and the head of the institute's ambient intelligence group Mitja Luštrek.
"If the phone knows how we are moving, it can for instance provide useful advice related to transport, choose to reduce the number of notifications when we're driving, or switch the phone from ringing to vibrating mode when we're using the noisy metro," Luštrek illustrated.
The teams were using data collected over several months by researchers from the University of Sussex during their daily commutes.
While mentioning the possibility of marketing the solutions to companies developing mobile applications and phones, Luštrek explained that similar methods were already being used at the institute in several intelligent systems projects, including for applications in healthcare.
Related: In world first, Ljubljana hospital grows new nose on patient’s forearm (video)
