Many workers and citizens have been forced to make a lifestyle change in the past two years due to the pandemic emergency. In order to keep a high level of personal health, the doctors suggest to do fitness exercises. Before the pandemic it was possible to do these exercises at the gym or during dedicated session in the office supervised by professional trainers. During the pandemic emergency the gyms were closed, the workers were forced to stay home and the people started to do gym exercises by themselves without the control of a professional figure. This situation could lead to several diseases associated to musculoskeletal disorders if the exercises are performed incorrectly. In this work, an approach based on the pose-estimator application OpenPose is developed. The reference exercise is an isometric squat performed by a professional trainer. During the exercise, thanks to a deep neural network, the pose-estimator gets a series of key-points and vectors which represent the user’s pose. A dataset of videos (for both the correct and incorrect postures) has been used to train several machine learning algorithms. The result is an automatic tool that recognizes incorrect poses during the exercise and helps the performer to correct it.
Posture Interactive Self Evaluation Algorithm Based on Computer Vision
Barberi E.
;Chillemi M.;Cucinotta F.;Milardi D.;Raffaele M.;Salmeri F.;Sfravara F.
2023-01-01
Abstract
Many workers and citizens have been forced to make a lifestyle change in the past two years due to the pandemic emergency. In order to keep a high level of personal health, the doctors suggest to do fitness exercises. Before the pandemic it was possible to do these exercises at the gym or during dedicated session in the office supervised by professional trainers. During the pandemic emergency the gyms were closed, the workers were forced to stay home and the people started to do gym exercises by themselves without the control of a professional figure. This situation could lead to several diseases associated to musculoskeletal disorders if the exercises are performed incorrectly. In this work, an approach based on the pose-estimator application OpenPose is developed. The reference exercise is an isometric squat performed by a professional trainer. During the exercise, thanks to a deep neural network, the pose-estimator gets a series of key-points and vectors which represent the user’s pose. A dataset of videos (for both the correct and incorrect postures) has been used to train several machine learning algorithms. The result is an automatic tool that recognizes incorrect poses during the exercise and helps the performer to correct it.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
