Functional electrical stimulation (FES) is a technology-based rehabilitation technique widely used following a spinal cord injury (SCI), stroke and other neurological pathologies, whereby short pulses of electric current are delivered to target muscles, generating contraction and consequent movement. Feedback is employed in FES systems to control the stimulation level to produce optimal movement patterns, including assessing muscle activity through electromyography (EMG). However, exogenous stimulation pollutes recorded EMG, making it challenging to assess volitional muscle activity and hindering the development of real-time neural controllers for FES. The overarching goal of this research was to develop and evaluate an FES system controlled by the underlying volitional muscle activity. To achieve this goal, we assessed common methods to extract the volitional EMG, designed a system for real-time FES control, and evaluated the effects of the control system on upper-limb movement. Multiple filtering methods for volitional EMG extraction have been previously proposed, however there is a lack of comprehensive evaluation and comparison between methods and, therefore, no consensus on which is the most effective to use. Here we compared the performance of the three most commonly used methods, high-pass, adaptive and comb filtering, by assessing the estimated volitional muscle activity across various conditions and performance measurements. The comb filter consistently provided the most accurate volitional EMG estimation across all conditions and performance outcomes, suggesting it is the most effective method for use in FES control systems. Furthermore, when designing an electrical stimulation system, consideration should be taken on how easily it can be controlled by the individual and how it affects their movement. We designed and tested two real-time FES systems controlled by the underlying volitional muscle activity, using a trigger and a proportional control scheme. II Able-bodied individuals were able to successfully control the stimulation using the designed systems, without having significant effects on how they activated their muscles and how smoothly they could perform the motion, by assessing EMG and motion data, validating the use of the FES systems. In conclusion, this work found, through comprehensive analyses, that comb filtering provided the best volitional muscle activity estimations during FES. The necessary EMG processing steps, including the comb filter, were successfully used to control the stimulation triggered by and proportional to the underlying volitional muscle activity in real-time. Both control systems could be used by able-bodied individuals without any negative effects on their ability to perform upper-limb movements. The designed software for this thesis can increase patient engagement and accommodation of residual volitional motor control during SCI rehabilitation, with immediate future applications including the incorporation of other promising rehabilitation techniques such as robotic assistance.