In current functional neuromuscular stimulation systems (FNS), control and feedback signals are usually provided by external sensors and switches, which pose problems such as donning and calibration time, cosmesis, and mechanical vulnerability. Artificial sensors are difficult to build and are insufficiently biocompatible and reliable for implantation. With the advent of methods for electrical interfacing with nerves and muscles, natural sensors are being considered as an alternative source of feedback and command signals for FNS. Decision making methods for higher level control can perform equally well with natural or artificial sensors. Recording nerve cuff electrodes have been developed and tested in animals and demonstrated to be feasible in humans for control of dorsiflexion in foot-drop and grasp in quadriplegia. Electromyographic signals, being one thousand times larger than electroneurograms, are easier to measure but have not been able to provide reliable indicators (e.g., of muscle fatigue) that would be useful in FNS systems. Animal studies have shown that information about the shape and movement of arm trajectories can be extracted from brain cortical activity, suggesting that FNS may ultimately be directly controllable from the central nervous system.