this study investigates the joint moment estimation performance and clinical application in studying muscle functions of the patient-specific EMG-driven neuromuscular model (PENm) for cerebral palsy patients. The PENm is a modified hill-type model designed for robotic application. The PENm evaluation employs the gait analysis technique using three dimensional motion capture system combined with EMG recording devices and the musculoskeletal modelling technique. The results show that the PENm could predicted joint moment based only on two EMG channels with an acceptable accuracy. In addition, the results also provide in-depth information via individual muscle forces on muscle functions for cerebral palsy patients with crouch gait. The design of gait rehabilitation robots and investigation of muscle functions for cerebral palsy patients can benefit from the application of the human internal state provided by the PENm.