INESC Key Detection

The INESC Key Detection (IKD) is a system for estimating the global key from musical audio and symbolic notation on the Tonal Interval Space. It incorporates a method for dynamically favouring either the major or minor key modes using a spatial displacement of beat-synchronous Tonal Interval Vectors (TIVs). A performance evaluation of the system has shown that it outperforms related key detection systems as well as the effectiveness of the mode bias — allowing users to fine tune this variable to improve correct key estimates on style-specific music datasets or balance predictions across key modes on unknown input sources.

An early version of the IKD system was submitted to the MIREX 2016 — Audio detection contest. Results can be accessed here. Although the algorithm didn't rank first in any of the evaluated datasets (MIREX 2005 Dataset and GiantSteps), when combining the two datasets the systems performs best out of the four submissions.