Gait pattern affects the quality of a walking robot. In this paper, an optimum human-like gait generation method is proposed for a seven-link biped robot. This method utilizes kinematic constraints based on human-like motion. At first, the constraints are parameterized with unknown parameters. Then, an optimization problem is defined with an objective function and the kinematic constraints. The objective function includes the energy consumption and walking stability. Unknown parameters appearing in the constraints are determined by solving the optimization problem. Joint variables are obtained by integrating a velocity form of the kinematic constraints. This approach is carried out for gait planning of the seven-link biped robot and its results are presented. This method allows studying some effective parameters on the gait such as constant knee angle duration, and torso inclination is optimized for different speeds. Our studies indicate that as walking speed increases the knee should extend and the torso should lean forward further in order to maintain better stability. Moreover, energy consumption increases as the torso leans forward.