Automated exercise evaluation and movement evolution

//Automated exercise evaluation and movement evolution

Frequent physical exercise provides numerous health benefits to individuals, such as improving endurance, fitness and muscle strength. Exercise can also reduce the risk of heart disease, type 2 diabetes, depression and dementia [1]. Especially for people with Parkinson’s disease, exercise plays an important role in maintaining a good level of mobility, posture and balance [2], which will allow them to be as autonomous as possible. Besides the physical benefits, exercise also has a positive impact on the general sense of one’s well-being [1], which is equally important for people with Parkinson’s disease.

Rehabilitation exercises for Parkinson’s patients can be performed either in private or in group sessions that take place in specialized centers, such as the Madrid Parkinson Association (APM) [4]. During these sessions, a physiotherapist along with an individual or a small group of patients perform a set of exercises that are suitable to the specific patients’ needs. Several repetitions of each exercise are carried out before moving on to the next exercise. In order to provide the physiotherapists with an automated and objective evaluation of the patients’ performance during these sessions, the Exercise Evaluation tool has been developed. Additionally, by keeping a record of these evaluations for long periods of time, months or even years, it is possible to observe the evolution of one’s physical condition.

As can be seen in figure 1, the Exercise Evaluation tool receives input from a Kinect sensor and displays at the top part of the window a view of the room and the detected persons. In order to better preserve the privacy of the people performing the exercises, the Kinect depth image is displayed instead of a regular color image, which provides less identifiable information. At the bottom part of the window, the evaluation score of each detected person for the selected exercise is shown, both in numeric form and as a progress bar. To produce the evaluation scores, the tool has a number of trained models, one for each pre-defined exercise, which continually evaluate the input from Kinect. Although the user interface displays the evaluation of the model that corresponds to the selected exercise, the evaluations of all models are gathered and saved for further analysis.

table 1

Figure 1. The Exercise Evaluation tool.

The tool interface provides the user with an immediate feedback regarding the performed exercise and how well it is being executed, at each instance, by up to six people. At the end of the session, however, after all model evaluations have been gathered, we have the opportunity to further analyze the data and gain a deeper insight about the performance of each tracked person. Figure 2 shows the analysis results for data gathered while a person was performing the exercise Upper Limb & Opposite Lower Limb Flexion. As a first step, the algorithm tries to find distinct repetitions for each detected person and for each exercise. Each distinct repetition is marked by three red dots: one at the 25% mark (with ascending trend), one at the top and one at the 25% mark again (with descending trend). In this recording, only a single person was tracked by Kinect. Then, by taking into account the distinct repetitions and scores from each exercise model, the system predicts the exercise being performed, which in this case was Upper Limb & Opposite Lower Limb Flexion. Finally, the repetitions for the predicted exercise are counted and their scores are averaged to produce one final score.

table 2

Figure 2. Analysis results.

A long-term pattern, that describes the movement evolution of a person through time, can be observed using a summary of one’s scores for a specific exercise for extended periods of time. As an example, figure 3 illustrates a hypothetical scenario, where the weekly scores of a person are averaged into monthly scores for a period of two years. The figure shows how these monthly scores evolve over time, giving the physiotherapist an objective insight about how the physical condition of a patient is advancing. This information, together with the physiotherapist’s own opinion about the progression of one’s physical condition, may inspire changes to the exercise protocol a patient is following.

table 3

Figure 3. Movement evolution.






Dr. Vassilis Solachidis
Information Technologies Institute
CERTH – Centre for Research and Technology Hellas

Use of cookies

This website uses cookies to improve your experience. By continuing to browse the site you are agreeing to our use of cookies. ACCEPT