Abstract: In this paper a virtual bass system based on audio source separation is proposed. The process of traditional virtual bass system is analysis and we found that the bass performance of different frequency components is not the same. That impact the performance of bass enhancement greatly. In order to save the distortion and improve the bass effect we propose a new virtual bass system based on audio source separation. We separate harmonic and percussive component apart before virtual bass generated. Experiment shows the system we proposed has better bass effect and quality with less computation time compare with the traditional method.