We present a study on the automatic classification of live and studio audio recordings, an important meta-information for music catalogue browsing and music recommendation systems. Several possible input representations (MFCCs, Mel spectrograms, VGGish) are combined with the classifiers GMM, SVM, and CNN to identify the most powerful approach. The results show that a CNN with VGGish input clearly outperforms other approaches and that its detection accuracy is high enough to be useful in practical applications.

Authors: Manjunath, Tejas; Pawani, Jeet Kiran; Lerch, Alexander
Affiliation: Georgia Institute of Technology, Atlanta, GA, USA
AES Convention: 149 (October 2020) Paper Number: 10399
Publication Date: October 22, 2020 Import into BibTeX
Subject: Audio Processing
Permalink: https://www.aes.org/e-lib/browse.cfm?elib=20936

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.

Learn more about the AES E-Library

E-Library Location:   /conv/149/10399.pdf

Start a discussion about this paper!