To achieve the balanced walking of humanoid, it is necessary to estimate dynamic stability of the system. However, in a small size humanoid which has restricted system resources, it is hard to satisfy the performance level desired by dynamic analysis such as on-line ZMP estimation. Therefore, in this paper, we proposed the feasible method of gait pattern generation and stabilization based on non-linear oscillators and sensory feedback which does not require high system performance. The suggested method was tested on the open humanoid platform DARwIn-OP for evaluation, and the result showed that the gait pattern can be generated in 0.008 sec with retaining of walking balance.