Eco-socio-botics 2022: Social Robotics for sustainability
Workshop, 16th December 2022
organized by
Ilaria Alfieri, Antonio Fleres, Luisa Damiano
A satellite workshop of the International Conference of Social Robotics 2022 (ICSR2022), 13-16 December 2022
Talk Details:
Towards a personalized interactive robotic fitness instructor for the elderly
Apostolos Kazakis1 and Chairi Kiourt2
1Department of Computer Science, School of Science, International Hellenic University, Kavala, Greece
2Athena-Research and Innovation Center in Information, Communication and Knowledge Technologies, Xanthi, Greece
Abstract: The average age of the world's population is increasing rapidly. At the same time, the benefits of regular exercise for the elderly are vitally important, considered by health scientists. However, the cost of a personal session with a fitness coach or physical therapist is quite high. So, we believe that the use of technology can provide solutions that in the long run are very likely to be proven very useful to the average person and especially the elderly making their personal training and rehabilitation very feasible. For this reason, we propose a personalized interactive AI-driven robotic fitness instructor for the elderly focusing on their peculiarities.
Fasola & Matarić, 2013 designed a Socially Assistive Robot (SAR) system to engage elderly users in physical exercise. Elderly was motivated and guided by the robot to perform arm gesture exercises. In (Görer et al., 2017), an autonomous robotic exercise tutor for elderly people was proposed operating in two different phases. In the first phase, the robot learns the exercise motion by a human, while in the second phase the robot operates as a trainer for the elderly person. Furthermore, in both studies, authors conclude that their proposed systems were evaluated as useful by the elderly participants and their feedback was quite positive.
In this study we are using a humanoid robot as a fitness instructor, which demonstrates exercises to elderly and simultaneously evaluates their performance through observation (computer vision). In addition, we are focusing on the development personalization approach focusing on more efficient instruction. Our system provides personalization approaches both at the level of the exercise program and at the level of individual performance. The robot recognizes the person through computer vision approaches, so as to retrieve knowledge about their status and physical condition. At the same time, it matches the person with their personal workout plan and retrieves their performance in each exercise from a database. The robot demonstrates the exercise to the elderly and monitors their imitation approaches and intervenes where necessary with further instructions.
It continuously estimates the body posture through computer vision and machine learning approaches of the trainee and compares it with the ideal body posture they should have while performing the exercise. When the trainee's body posture is not appropriate, the robot produces and addresses meaningful verbal instructions, such as "Lift your right hand", to correct them. At the same time, it motivates the elderly to improve their performance in the exercise they are performing.
Technically, the system consists of an Aldebaran NAO robot that communicates with a database, which contains all the necessary information related to the workout programs and performance of the elderly. Moreover, various Computer Vision and Machine Learning approaches related to pose estimation and face recognition are being exploited. For the needs of this study, at this stage, two minimal workout programs were developed and proposed by a specialist personal trainer. Each workout plan contains three exercises.
Acknowledgments
This work is supported by the MPhil program "Advanced Technologies in Informatics and Computers", hosted by the Department of Computer Science, International Hellenic University.
References
Fasola, J., & Matarić, M. J. (2013). A socially assistive robot exercise coach for the elderly. Journal of Human-Robot Interaction, 2(2). https://doi.org/10.5898/JHRI.2.2.FASOLA
Görer, B., Salah, A. A., & Akın, H. L. (2017). An autonomous robotic exercise tutor for elderly people. Autonomous Robots, 41(3), 657–678. https://doi.org/10.1007/S10514-016-9598-5/FIGURES/12
