Abstract:

Weightlifting exercises are not easy to get the hang of, especially for beginners. Weightlifters need to be well informed and ready before they start working out with dumbbells, or otherwise, they might end up with very problematic injuries. In this paper, building a model that allows the detection of wrong movements during the workout so that users can quickly check it and correct their mistakes using machine learning and deep learning techniques.   It presents the possibility of having a model that can guide people throughout their weightlifting process.  This model will massively help reducing injuries by tracking body motion with no requirements needed other than the Kinect camera.  The results of testing showed the difference in performance by three different algorithms k-nearest neighbors algorithm (KNN), the support vector machine (SVM) and the FAST Dynamic time warping  (FAST-DTW)  dynamic time warping concluding that the  FAST-DTW  was the most efficient one with higher accuracy and faster processing time.