Functional electrical stimulation (FES) is a useful rehabilitation technique for restoring motor capability of spinal cord injured (SCI) patients by artificially driving muscle contractions. Real-time FES systems with online modulation ability are in great need for clinical applications. In this work, a system for real-time estimation of joint torque with evoked electromyography (eEMG) is presented. Kalman filter (KF) is adopted and embedded into the system as the online torque estimator. The real-time estimation system would be promising toward FES control with consideration of torque changes caused by muscle fatigue.