A significant problem associated with active noise control (ANC) systems for home windows is that they use direct microphone measurements for reference and error signals. When other sound besides external noise (e.g. music or speech) exists inside a home, the reference and error microphone measurements are "polluted". Using the "polluted" measurements as reference and error signals has adverse effects on the ANC system. This paper proposes to integrate the ANC system with a wave separation algorithm, which separates the "unpolluted" reference and error signals from the direct microphone measurements. The performance of the resulting ANC system with wave separation is experimentally tested in a cabin equipped with a window and results are presented. Experimental results show that the new system is able to preserve desired internal sound while cancelling uncorrelated external noise.