An alternative to inverse dynamics joint torques estimation in standing human is proposed by the use of an unknown input observer that allows the real time estimation of both the joint torques and the angular velocities from angular positions measurement. To design such observer, a state space descriptor dynamical model of the considered movement is proposed. Then, the considered model being non linear, we used Takagi-Sugeno fuzzy modeling to synthesis the non linear observer and to prove the convergence of the state (position, velocities and torques) estimation errors.