Observado

Introduction

This is a simple GUI tool for chord detection developed from the example in librosa.sequence.viterbi_discriminative, with the ability to detect some seventh chords by using the k-NN algorithm.

The main idea is using the k-NN algorithm to tell some triads and sevenths chords apart after the result of librosa.sequence.viterbi_discriminative. Chroma features of audio files generated by MIDI chord templates are used as the dataset.

This is only a project for my coursework which does not guarantee the accuracy. It would be better not to use it to do chord detection in practice.

Requirements

• Timidity++
• Qt

Python

• joblib
• librosa
• matplotlib
• numpy
• pandas
• scikit_learn
• scipy
• tqdm

Setup

First, install Python requirements.

pip install -r requirements.txt

Then generate data. It may take a while.

cd observado/
python generate.py

You may directly use main.py, or build the GUI tool with CMake.