This paper presents a novel visual motion estimation technique for tracking articulated human body configurations in complex video sequences. The method uses the product of exponential maps and twist motions, integrated into a differential motion estimation framework. This approach allows for solving simple linear systems, enabling robust recovery of kinematic degrees of freedom in noisy and self-occluded scenarios. The technique is demonstrated on video sequences of people performing full-body movements and on historical motion studies by Eadweard Muybridge. The system successfully recovers and re-animates these motions with high accuracy. The paper introduces a motion tracking framework that incorporates kinematic chain constraints using twist and product of exponential maps. This formulation simplifies motion estimation and leads to robust tracking results. The method is extended to handle multiple camera views, improving the robustness of angular parameter estimation. An adaptive support map using EM is introduced to better align the motion model with the actual body shape. The tracking algorithm is applied to both single camera and multi-camera video sequences. It successfully tracks human motion in frontoparallel and oblique views, as well as in historical Muybridge motion studies. The results show accurate tracking of joint angles and body poses, even in challenging conditions such as occlusions and varying scales. The method is also used to re-animate historical motion data, demonstrating its effectiveness in recovering complex human motions. The paper concludes that the proposed technique, based on the product of exponential maps and twist motions, provides a robust and efficient solution for articulated visual motion tracking. Future work will focus on handling large motions, such as those in high-speed running sequences. The method is a differential approach, which may struggle with large frame-to-frame motions, but can be improved with an initial coarse search stage. The research is supported by various institutions and acknowledges contributions from colleagues and collaborators.This paper presents a novel visual motion estimation technique for tracking articulated human body configurations in complex video sequences. The method uses the product of exponential maps and twist motions, integrated into a differential motion estimation framework. This approach allows for solving simple linear systems, enabling robust recovery of kinematic degrees of freedom in noisy and self-occluded scenarios. The technique is demonstrated on video sequences of people performing full-body movements and on historical motion studies by Eadweard Muybridge. The system successfully recovers and re-animates these motions with high accuracy. The paper introduces a motion tracking framework that incorporates kinematic chain constraints using twist and product of exponential maps. This formulation simplifies motion estimation and leads to robust tracking results. The method is extended to handle multiple camera views, improving the robustness of angular parameter estimation. An adaptive support map using EM is introduced to better align the motion model with the actual body shape. The tracking algorithm is applied to both single camera and multi-camera video sequences. It successfully tracks human motion in frontoparallel and oblique views, as well as in historical Muybridge motion studies. The results show accurate tracking of joint angles and body poses, even in challenging conditions such as occlusions and varying scales. The method is also used to re-animate historical motion data, demonstrating its effectiveness in recovering complex human motions. The paper concludes that the proposed technique, based on the product of exponential maps and twist motions, provides a robust and efficient solution for articulated visual motion tracking. Future work will focus on handling large motions, such as those in high-speed running sequences. The method is a differential approach, which may struggle with large frame-to-frame motions, but can be improved with an initial coarse search stage. The research is supported by various institutions and acknowledges contributions from colleagues and collaborators.