Knowing the number of sources present in a mixture is useful for many computer audition problems such as polyphonic music transcription, source separation, and speech enhancement. Most existing algorithms for these applications require the user to provide this number thereby limiting the possibility of complete automatization. In this paper we explore a few probabilistic and machine learning approaches for an autonomous source number estimation. We then propose an implementation of a multi-class classification method using convolutional neural networks for musical polyphony estimation. In addition, we use these results to improve the performance of an instrument classifier based on the same dataset. Our final classification results for both the networks, prove that this method is a promising starting point for further advancements in unsupervised source counting and separation algorithms for music and speech.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.
The Engineering Briefs at this Convention were selected on the basis of a submitted synopsis, ensuring that they are of interest to AES members, and are not overly commercial. These briefs have been reproduced from the authors' advance manuscripts, without editing, corrections, or consideration by the Review Board. The AES takes no responsibility for their contents. Paper copies are not available, but any member can freely access these briefs. Members are encouraged to provide comments that enhance their usefulness.