A Video Analysis System Using K-Nearest Neighbour Algorithm for Error Detection in Squat Form
Lakshay Bansal, Vinay Vishwakarma · 2023
In this paper, a video analysis system is developed using K-Nearest Neighbour (KNN) algorithm to analyze the accuracy of physiotherapy exercises. This system captures videos of patients undergoing physiotherapy treatments. The K-Nearest Neighbour (KNN) algorithm is used to assess the video and identify any deviations from proper during the physiotherapy treatments. To develop this configuration, a dataset of physiotherapy exercises performed by patients was collected by trained physiotherapists. This dataset is used to train the KNN algorithm, enabling it to optimize its performance by selecting the optimal number of neighbours and distance metric. The performance was evaluated using a separate dataset of physiotherapy exercises. the results show that the KNN algorithm accurately detected errors in the patients' exercise techniques. This study achieved a notable accuracy rate of 92% in detecting errors, thus highlighting the effectiveness of the proposed approach. Particularly, the video analysis system is designed to detect errors in the squat form, such as incorrect knee and hip alignment or improper movement. To accomplish this, a dataset of patients performing squats was collected and the KNN algorithm was trained on this dataset to detect deviations from the correct form. This present study provides a practical demonstration of the proposed video analysis system application in the context of physiotherapy treatment using the KNN algorithm. The main objective of this study is to the analysis of squat exercises and assess their proper execution.