To perform the motion control of a 12 DOFs lower limb exoskeleton for rehabilitation and power augmentation (LLE-RePA), this paper proposes a linear discrete-time extended state observer (LDESO) based intelligent PD (iPD) controller with sigmoid function-based tracking differentiator (STD). A three-order discrete-time STD is designed to obtain smooth real-time reference velocity and acceleration signal. The adopted iPD controller is a kind of model free method based on ultra-local model, which formulates complex human-exoskeleton dynamics in a simple way and make closed-loop PD parameters easy to be tuned. The LDESO is utilized to estimate and eliminate the lumped total disturbance in ultra-local model which contains unknown model dynamics and unpredictable human-exoskeleton interaction influence. The estimation result of LDESO is used to compensate the control input. Simulation and experimental results with exoskeleton LLE-RePA demonstrate the high performance of the proposed method.