Highly dexterous robotic hand prostheses are becoming commercially available, leading to a significant need for intuitive methods to control such devices. Using surface EMG (electromyography) recordings, pattern recognition based myoelectric prostheses offer intuitive control but at the cost of increased training time. In addition, such approaches often ignore the underlying biomechanics and sensory information required for dexterous hand movements. Biomimetic models provide an intuitive control scheme by embedding the mechanical and neural coupling that naturally exist in the human hand. Here, we combined biomimetic models with the high channel-count intramuscular myoelectric recording systems to overcome these challenges. Our approach offers an alternative that largely eliminates the need for training required for such complex control systems. Combined with the sensory feedback technique developed in our group, this motor decoding technique has shown a great potential for enhancing control of advanced prosthetic hands.