This paper presents a method for adaptation of the Human Motion Capture Data (HMCD) for walking motion of a biped robot. We developed a dynamics filter, that generates a physically consistent walking motion from references trajectories obtained by HMCD. The method uses the inverse dynamics of the robot and the concept of the Zero Moment Point for the stability. A scheme for mapping the human trajectories of a gait cycle to the desired joint trajectories of the robot, through periodic joint motion is presented. The resulting trajectories are parametrized by Fourier series expansion, this approximation is important to preserve the naturality of the walking motion. Finally, we tested the proposed approach on a humanoid robot.