The inertial parameters are important to generate motion patterns for humanoid robots. Conventional identification methods can be used to estimate these parameters; however they required the joint torque estimates that can be obtained by modeling of the transmission or by direct measurements. To overcome that issue we have recently developed a new method to estimate the inertial parameters of legged systems. By using the base-link equations only, we obtain a reduced identification model that is free of joint torque estimates. In this paper we propose to apply the method to a human-size humanoid robot. The preliminary experimental results are given and discussed.