利用核依賴估計來進行多軌自動混音

摘要

Due to the revolution of digital music, People can create acceptable quality recordings in a home studio with cheaper gear. But after multi-track recording, they will next do the essential part of music production: Mixing which combine multi-track recording into one or more channel. The Learning curve of how to mix is very high due to some sound engineering and psychoacoustic background knowledge. It is difficult to get good mixdown for non-specialist in sound engineer. In this paper, we want to use a supervised learning method for automatically mixing multi-track recording into coherent and well-balanced piece. In our system, due to lack of data source of mixing parameter, first we estimate the weight of mixing parameter by using the relation between raw multi-track and mixdown. Then using estimated parameter, we adopt kernel decency estimation method to create our mixing model based on the dependency between tracks.