Machine learning for gait analysis and performance prediction. (GaitMaven)

Main funder

Research Council of Finland

Funder's project number: 323473

Funds granted by main funder (€)

175 522,00

Funding program

Academy Project, AoF (Research Council of Finland)

Project timetable

Project start date: 01/01/2020

Project end date: 31/08/2023

Summary

This project’s aim is to enable kinesiology research, which nowadays has to be done in a well-equipped laboratory, to be done in the field using a single body-mounted sensor unit integrating accelerometers, gyroscopes, compasses, barometers, a GPS receiver and a heart rate monitor. Along with the traditional metrics such as cadence, speed, step length and vertical oscillation, we will compute ground contact time, ground reaction forces, running power, running economy, lactate threshold and fitness level (VO2max). This equipment can replace expensive and cumbersome optical tracking systems, in-shoe pressure measurement systems, force plates, lactate analyzers and oxygen uptake measurement equipment. Another advantage of our approach compared to current methods is that natural movement is not impeded at the expense of measurement accuracy.

The rapid evolution of sensor cost, size, and energy efficiency has driven a boom in motion tracking applications in areas such as sports performance monitoring and clinical gait analysis. The new sensor technologies also enable research in physiology of sport and exercise in athletic and clinical settings. In particular, the use of very low cost sensors such as those in mobile phones can support the development of kinesiology science in developing countries. On the other side, development of advanced machine learning techniques such as deep learning allows indirect estimation of parameters based on only the sensor data and create surrogates for expensive measurement equipment.

The new technology developed in this project can be applied to different sports and human activities. In particular, the following areas will be considered: walking, running, motion disorder diagnosis and geriatric studies. The developed systems will be tested by real-time estimation of biomechanical and physiological parameters and compared with gold standards in biomechanics research. The sensor fusion and machine learning algorithms, and software that will be published in the project will help app programmers to take full advantage of the integration of sensors to deliver advanced performance analysis in sports and ambulatory monitoring.

The project builds on the expertise in machine learning and multi-sensor technology with applications to motion tracking and classification of the teams at Tampere University of Technology (TUT), and the expertise in biomechanical studies at University of Jyväskylä (JYU). The results of the project will be published in the key conferences and journals, widely respected in biomechanics, machine learning and sports technology communities worldwide, and made available for deployment into advanced R&D in academia and industry.