Physiotherapy-based human activity recognition using deep learning

Nowadays, continuous human activity recognition is being studied broadly by investigators for diverse applications. However, the studies based on physiotherapy action tracking from the physiotherapy video dataset are limited. Hence, the physiotherapy dataset has been considered in this present study. Moreover, deep learning-based (DL) neural networks have promoted the enhancement of activity detection study to become an essential technique. DL-based neural networks, such as long short-term memory, can automatically acquire the significant features from the physiotherapy video of sub-activity and main activity. Nevertheless, some physiotherapy videos are inappropriate and correspond to insignificant actions. Consequently, these insignificant actions can cause the recognition of continuous movements. Therefore, a novel strawberry-based recurrent neural framework is proposed to address this issue. Here, a physiotherapy video is taken as the input, and this dataset consists of several actions. Consequently, all the steps are done by one single person. So, the proposed design initially identifies all subactivities based on that sub-activities, and the main physiotherapy actions were classified. After that, repeated action counts and their starting and ending times are evaluated. Finally, the present study's design is considered in terms of performance metrics. Three things are mentioned in this article. First the class determines whether the human body is static, dynamic, or transitional, which class indicates the position of action. To recognize the main activity, it is important to first identify all subactivities in the physiotherapy video. Then, you should count the number of times each sub-activity was performed and how long it took overall. The proposed model was implemented using the Python platform, and the results were compared with the existing models. The proposed model shows higher recognition accuracy in comparison.