Content-driven automatic music equalization is becoming notably popular among the audio engineering society. The so far established methods are related to the genre and/or the use of a reference track. This work aims to investigate the relationship between the spectral energy of a frequency band and the equalization preference, for different genres. Towards this objective, five predefined equalization curves were evaluated through subjective tests, for ten music excerpts of various music genres. Initially, the statistical significance of linear regression fittings between single mel-band energy and the number of participants who opted for an equalization curve was assessed to determine the extent to which specific mel-band energy levels can describe the preference for an equalization curve. Subsequently, the significant bands were incorporated into a multiple linear regression model to explore the combinations of mel-bands that optimally predict the equalization preference. Experimental results have proven valuable linear relationships between the energy of specific frequency bands and equalization preference, particularly for the low-frequency boosting, paving the way for automatic equalization algorithms improvements.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=22539
(1347KB)
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 Join the AES. If you need to check your member status, login to the Member Portal.
