The biologically-inspired motion control theory was introduced to ameliorate human-robot interaction flexibility and multi-joint autonomy in walking assist. A new type of human-robot interaction (HRI)-based coordination control method which combined a hip CPG (central pattern generator) control, a knee hierarchical impedance control and a hip-knee linkage control was explored. Firstly, the characteristics of CPGs self-oscillation and external communication were utilized to obtain the ideal master/slave hip joint trajectories, and CPG-units symmetrical inhibition were established to maintain anti-phase of the left and right hip joints. Secondly, according to the requirements of gait, a high and low hierarchical impedance control law was designed, and thus the knee joint torques during support stage and swing stage were obtained respectively. Finally, a hip-knee joints linkage mechanism was built to incorporate the hip joint CPG control with the knee-joint impedance control, and thus natural and relevant hip-knee joints trajectory was realized. The stability of the HRI-based coordination control method was researched by using Lyapunov stability theory. Walking experiments were carried out to certificate that the proposed control method was effective in generating nature hip-knee joints trajectory in walking.