This study explores the use of MediaPipe for markerless joint angle estimation and its application in calculating joint moments during sit-to-stand movements. The problem is the absence of an accurate, affordable, and easy alternative to costly, complex marker-based systems that restrict natural movement. Our approach leverages a rapid setup using a single camera to capture motion, making it a cost-effective and accessible solution for motion analysis. We conducted experiments with 15 healthy volunteers, recording their STS movements using a smartphone camera. The videos were processed using MediaPipe to estimate joint angles at the ankle, knee, and hip. These angles were then used to compute angular velocities, accelerations, and joint moments through an inverse dynamics model. We compared the performance of our markerless approach with conventional marker-based systems and MATLAB simulations. The results showed strong consistency in hip and knee joint moment estimations, with correlation coefficients of 0.94 and 0.95, respectively. While the ankle joint demonstrated moderate agreement (R = 0.11), the findings confirm the method’s reliability for lower limb analysis. Our results also highlight the potential of our method for real-time applications in clinical and sports settings. The study underscores the advantages of using a markerless system, including ease of setup, reduced costs, and the elimination of movement restrictions associated with traditional marker-based systems. This research contributes to the field by providing a detailed analysis of STS movements using an innovative, deep learning-based approach. The findings support MediaPipe’s effectiveness for biomechanical analysis and its potential in clinical and rehabilitation settings.
Akturk et al. (Tue,) studied this question.