Functional electrical stimulation is one of effective and important means of pathological tremor suppression. However, due to the complex physiological and structural characteristics of the wrist musculoskeletal system, it is difficult to develop its accurate model, which greatly limits the performance of tremor suppression. In this paper, we first establish a Hammerstein model of wrist musculoskeletal system and propose a parameter identification procedure. Then, a feed-forward linearization controller is designed to cancel the nonlinear muscle recruitment characteristics. An active disturbance rejection control is proposed to estimate and compensate the model errors and external disturbances. The bandwidth method and stability margin analysis are used to complete the parameter tuning of active disturbance rejection controller and the tracking performance and disturbance rejection property are verified by simulation. Finally, based on the tremor suppression platform, comparative experiments with PID controller are carried out to verify the effectiveness of the proposed control method.