A multichannel active noise control system based on Empirical Mode Decomposition (EMD) is proposed in order to improve the noise reduction of the control system. The multiple noise source signals are decomposed adaptively and respectively by EMD at first, and loudness of all Intrinsic Mode Function (IMF) for each signal is calculated in such a system. Then the shaping filter of error signal is designed according to the amplitudes of IMFs’ loudness. The filter designed by this method can restrain the small-loudness components of noise source signals better than A-weighting shaping filter. The noise inside a car is measured, and the noise reduction of dual-channel active control systems is simulated. The results show that the system proposed by this paper can improve the subjective noise reduction in comparison to those with Filtered-X LMS and A-weighting shaping filter.