Real-time Deadlift Form Analysis Using Support Vector Machine Algorithm and Kinect V2 Sensor

We developed a real-time program utilizing the Support Vector Machine (SVM) algorithm in conjunction with the Microsoft Kinect V2 sensor to analyze the proper execution of the conventional deadlift exercise. A user-defined system was developed to detect movements using the Support Vector Machine algorithm in programming languages and assess the form of the deadlift. We calculated the posture angles based on Chaffin’s model to estimate the errors from the user’s form in real-time and display numerical values regarding the errors. The results of the system were evaluated based on a professional trainer’s analysis. The system was validated by comparing it with five experienced trainers, and the outcomes demonstrated an impressive accuracy rate of 90.54% in correctly identifying the proper form of the conventional deadlift exercise. The proposed system effectively discerned correct and incorrect forms of the conventional deadlift in real-time, leveraging the SVM Image Processing Algorithm and Kinect V2 Sensor. Future research is necessary to overcome the study’s limitations and enhance its accuracy.