Performing prescribed physical exercises during home-based rehabilitation programs plays an important role in regaining muscle strength and improving balance for people with different physical disabilities. However, patients attending these programs are not able to assess their action performance in the absence of a medical expert. Recently, vision-based sensors have been deployed in the activity monitoring domain. They are capable of capturing accurate skeleton data. Furthermore, there have been significant advancements in Computer Vision (CV) and Deep Learning (DL) methodologies. These factors have promoted the solutions for designing automatic patient's activity monitoring models. Then, improving such systems' performance to assist patients and physiotherapists has attracted wide interest of the research community. This paper provides a comprehensive and up-to-date literature review on different stages of skeleton data acquisition processes for the aim of physio exercise monitoring. Then, the previously reported Artificial Intelligence (AI) - based methodologies for skeleton data analysis will be reviewed. In particular, feature learning from skeleton data, evaluation, and feedback generation for the purpose of rehabilitation monitoring will be studied. Furthermore, the associated challenges to these processes will be reviewed. Finally, the paper puts forward several suggestions for future research directions in this area.

Fine-tuning the hyperparameters of the Support Vector Machine (SVM) is crucial to making it an accurate classifier in fault detection. The convergence rate of the optimiser and its simplicity are also vital for achieving the optimal boundary. Therefore, a highly accurate fault detection algorithm that combines SVM with a simple, easy-to-implement, and stable optimiser called Projection Recurrent Neural Network (PRNN) is proposed in this paper. The algorithm focuses on detecting bearing faults in induction motors using experimentally measured stator currents. The initial dataset is pre-processed using Discrete Wavelet Transform (DWT), Power Spectral Density (PSD), and cepstrum analysis. A feature set is then derived from several statistical metrics, and Kernel Principal Component Analysis (KPCA) is used to select the most salient features. The SVM is trained using these discriminative features, and its optimisation problem is reformulated as Constrained Nonlinear Programming (CNP). The PRNN is introduced to solve the CNP to effectively determine the optimal decision boundary of the SVM. The study evaluates the accuracy of the proposed algorithm and highlights the advantages of using PRNN in SVM, its dependence on the type of kernel function and the selected features. A 4 kW induction motor is used as a prototype machine for experimental data collection under healthy and faulty conditions. The results show that the proposed algorithm leads to more accurate fault detection compared to the conventional SVM and recent fault detection methods.