This study shows an active vibration control using a phase and an amplitude of a mean field of oscillators. Recently, we have proposed an active mass damper (AMD) system using a neural oscillator, which can flexibly synchronize with a target structure’s vibration response. In the proposed system, a neural system including a single oscillator and a position controller form the controller: the neural system generates a target path of an auxiliary mass of the AMD and then, the auxiliary mass is position controlled to the target path to absorb the vibration energy of the structure. However, the oscillator could unfortunately include a non-linear response when the frequency of the periodical input is dramatically different from the oscillator’s eigen frequency. To suppress the undesirable single component of the control signal, this study will use the phase and the amplitude information of an arithmetic mean of the oscillators for the target path of the auxiliary mass of the AMD. Cleaning the control signal could improve an input-output stability of the proposed system and enable to vibration control a specific vibration mode of a multi-degree-of-freedom system such as high-rise buildings.