Online trainer corrects individually tailored movements

Since the pandemic, working out at home in front of the computer has become a habit for many. But are these exercises, be they for rehabilitation or athletic purposes, being performed correctly? Roman Dorrek is developing software that can identify whether individual movements are being performed in the most beneficial way possible.

It is commonly believed that it does not matter whether movements and exercises are performed perfectly, as long as you are engaging in physical activity. Roman Dorrek, a data-focused sports scientist and pre-doctoral researcher in the Ada Lovelace Programme, emphatically disagrees: “You definitely shouldn’t be doing that if you’ve had an injury and want to get your body back into shape with physical therapy or training. A healthy body can handle a lot, but there are some movements that are good for you and others that aren’t, and that very much depends on your individual needs.”

To ensure that patients who are in rehabilitation, as well as other people who are exercising, perform the movements correctly and in a way that is appropriate for them, Roman Dorrek is developing an online trainer or online physiotherapist as part of his doctoral project. Using individually configured settings and cameras, the computer recognises whether exercises are being performed correctly and where there is room for improvement. Compared to existing systems, this system should offer improved validity and no longer depend on a catalogue of movements, as Roman Dorrek explains: “Many existing solutions require you to specify in advance that you are going to perform a squat. The system then corrects your knee position. We want to develop our tool in a more holistic way and analyse motion as a whole, regardless of predefined movements.”

In addition, most systems are designed to take absolute joint angles as a reference point. “There is, of course, a perfect technique, but it is very individual,” Roman Dorrek goes on to explain. For example, there are anatomical differences between men and women that also affect the angles used in certain movements. Previous injuries also influence how an exercise can be performed most effectively.

To ensure that the proposed system can learn as much as possible about individual differences and appropriate movements, the tool will incorporate Artificial Intelligence. The project is therefore interdisciplinary in nature; Roman Dorrek is supervised by Jan Wilke (Department of Movement Sciences) and Gerhard Friedrich (Department of Artificial Intelligence and Cybersecurity). For Roman Dorrek, immersing himself in computer science is both exciting and challenging. But technical topics have been part of his education all along: He studied sports equipment technology at the University of Applied Sciences Technikum Vienna, graduating with a Bachelor’s degree, and then completed a Master’s degree programme in Performance Analysis of Sports at Otto von Guericke University Magdeburg. He moved to Klagenfurt in December 2023 in response to an advertisement for pre-doctoral positions in the Ada Lovelace Programme. For the past six months, he has been collecting data from people training in the sports science laboratory on the campus of the University of Klagenfurt: “To do this, the test subjects are covered with reflective markers. When they perform movements, these are recorded by a video camera system that can accurately capture the positions of body parts such as arms and legs at any given moment.”

The new tool isn’t just for people who want to work out at home, but it’ll also be a big help for physical therapists. “Physiotherapy sessions are usually only held once a week or every two weeks. If private training sessions at home are recorded using the tool and the exercises are corrected, the patient stands to benefit. This also provides valuable feedback for the therapist on the progress made between therapy sessions,” Roman Dorrek explains.