One of major disabilities that occur because of war or incidents is loss of limbs. Using of artificial prosthesis has a long history. After World War II, many researches have been accomplished to add movement abilities to artificial limbs. For this purpose, biological signals were considered as a media between man and prosthesis. One of these signals is electromyogram or EMG signal. This signal can only be used for control of the prosthesis, if it is already analyzed by pattern recognition methods to determine desired movement and produce proper control signals. In this research, biological generation of EMG signals, their properties, and different methods in pattern recognition for movement class extraction will be introduced. The main goal of this research is a realistic assessment of practical problems that exist in implementation of cybernetic hand. For this purpose, MLP classifier and IAV of signals are used for pattern recognition phase. One of the main problems that exist in current cybernetic hands is their control systems. In these methods, after determination of movement class, prosthesis starts to move with a predetermined and constant velocity. This makes the movements unnatural. Furthermore, we can not add combinational movements to prosthesis. To solve these problems, we closely observe the natural system behavior in order to find ideas that make them significantly perfect and efficient. For this purpose, we have modeled the musculoskeletal system of arm and simulate brain’s control on it. The results of this modeling lead us to introduce a new control topology to control arm movements more naturally than before. Finally, the proposed control method will be simulated and results are analyzed