SRNF-PAR: Strawberry and Recurrent Neural Framework-based Physiotherapy Activity Recognition
Disha Deotale, Madhushi Verma, P. Suresh, Arpit Bharadwaj · 2024
Person activity recognition is now a research area in computer vision. The various applications of this area include healthcare, sports and suspicious activity detection. This study focused on physiotherapy exercise activity. The video dataset of physiotherapy is limited; therefore, it was collected from YouTube for shoulder therapy. The exercise has many activities thing is available to understand it all, here more than one activity in sequence wants to do with some activity repetition. Therefore, the deep learning-based algorithm such as in recurrent neural networks the long short-term memory technique will help to determine the exercise with the help of sub-activity order and the other parameter will identify which activity is repeated and the duration of exercise activity with the pose of the person to perform the activity such as static, dynamic and transition. All parameters were obtained using the novel approach of a strawberry-based recurrent neural network for physiotherapy activity recognition (SBRNN-PAR). This algorithm will classify the sub-activity, recognize the exercise with repeated counts and identify the human pose to perform the activity. The idea for a model shows a higher performance in evaluation.