In order for an active exoskeleton and its user to achieve synchronous motion, the intended motion of the user needs to be detected with enough lead-time to process and move the exoskeleton accordingly. This must also happen with a level of accuracy such that the exoskeleton does not impedes the motion of the user. Synchronous motion is difficult to achieve because human musculoskeletal motion is extremely complex with multiple muscles controlling multiple degrees of freedom of the joints. One promising method of reading human motion intent is with the detectible electrical signal that results from muscle activation measured via electromyography. This signal can be measured noninvasively on the surface of the skin, and is detectable approximately 100 ms before movement ensues. For the work presented in this thesis, a control scheme that associates muscle activation to future knee flexion was developed using artificial neural network machine learning algorithms. Artificial neural networks are designed to function much like the human brain. Inversely to how the brain decides on the movement that the body will take and then tells the muscles to activate accordingly, the algorithms will read the muscle activation signal and make informed estimations of the joint angles that the brain is trying to achieve. This method was used to create a model for anticipating error versus prediction time. Furthermore, the method was used to assess the necessary inputs for an algorithm to make accurate knee flexion predictions on a user independent from the algorithm’s training data.