This work presents the design and development of a six-channel system for acquisition and conditioning of electromyographic signals collected in the upper limb. The main objective of the work is to create a system that can be used as rehabilitation and training instrument for potential users of myoelectric prostheses. Using software rendering, feature extractions, classifier training and design of the mechanical model of the human arm, with the running movements of flexion, extension, pronation and supination of the forearm and the grasp in a reality environment virtual, providing rehabilitation therapy to different patients.