With help training of the lower limbs rehabilitation robot, the hemiplegia patients can be helped effectively recover. Applicable control method plays an important part in performance of lower limbs rehabilitation robot. According to the preferred method, sEMG was collected from no necrosis and healthy muscle, then, the effective action signals which are extracted from the sEMG transit to Fuzzy-Neural network classifiers to identify the movements intention of paralyzed patients, and then the lower limbs rehabilitation robots can assist paralyzed patients to achieve their intent. The simulation results indicate that the Fuzzy-Neural network classifiers can identify the movements intention well, and control method of sEMG can satisfy the demand of lower limbs rehabilitation robot.