An approach to model and estimate human walking kinematics in real-time for Physical Human-Robot Interaction is presented. The human gait velocity along the forward and vertical direction of motion is modelled according to the Yoyo-model. We designed an Extended Kalman Filter (EKF) algorithm to estimate the frequency, bias and trigonometric state of a biased sinusoidal signal, from which the kinematic parameters of the Yoyo-model can be extracted. Quality and robustness of the estimation are improved by opportune filtering based on heuristics. The approach is successfully evaluated on a real dataset of walking humans, including complex trajectories and changing step frequency over time.