This paper describes a novel method for controlling a human-like musculoskeletal leg robot by implementing muscle coordination in human voluntary movement. To examine human motor control involving the ill-posed problem of redundant degrees-of-freedom (DOFs), we introduce the concept of the agonist-antagonist muscle pair ratio (A-A ratio) which represents the coordination of agonist-antagonist muscle pairs. Our concept enables us to extract kinematic information (instantaneous equilibrium posture) from the electromyography (EMG) signals. We show two patterns of muscle coordination extracted from the A-A ratio dataset during human pedaling, clarifying the motor functions of these patterns. Then, by transferring the muscle coordination of human pedaling to a human-like musculoskeletal leg robot with redundant DOFs of pneumatic artificial muscles, we enable the robot to achieve pedaling movement.