Welcome to DECIBEL’s documentation website!¶
DECIBEL is a new system for Automatic Chord Estimation (ACE) which exploits MIDI and tab files to improve audio ACE, thereby implicitly integrating musical knowledge.
Automatic Chord Estimation is a fundamental task in Music Information Retrieval (MIR), which has applications in both music performance and MIR research. The task consists of segmenting a music recording or score and assigning a chord label to each segment. Although it has been a task in the annual benchmarking evaluation MIREX for over 10 years, it is not yet a solved problem due to recent insights into the subjective nature of harmony perception.
Given an audio file and a set of MIDI and tab files corresponding to the same song, DECIBEL first estimates chord sequences from all files. For the audio file, it uses existing state-of-the-art audio ACE methods. MIDI files are first aligned to the audio, followed by a MIDI chord estimation step. Tab files are first parsed to untimed chord sequences and then aligned to the audio. In a final step, DECIBEL uses a data fusion method that integrates all estimated chord sequences into one final output sequence.
We tested DECIBEL on the Isophonics reference annotations. Our results show that DECIBEL improves all of the MIREX 2017, 2018, 2019 and 2020 ACE submissions, as well as a state-of-the-art commercial ACE method.
Try DECIBEL yourself¶
If you want to reproduce our results or try DECIBEL on your own data set, you can clone the source code from https://github.com/DaphneO/DECIBEL.
The documentation can be found on the current website.