Harmonic Intervals Audio Dataset

Read full text
Authors: Daria Shanurina, Roman S. Savitskyi, Тетяна Анатоліївна Вакалюк
Cited by: 1

Abstract

This dataset contains 8,236 WAV recordings of harmonic intervals for music information retrieval and machine learning research. Harmonic intervals are pairs of notes played simultaneously, representing the distance between two pitches. The dataset is split into two subsets: recorded-dataset — recorded acoustic samples synthetic-dataset — generated/synthesized samples across multiple instruments The dataset is organized in the following hierarchical structure: harmonic-intervals-audio-dataset/ ├── recorded-dataset/ │ ├── MF/ │ │ └── / │ │ ├── forte/ │ │ └── piano/ │ ├── NF/ │ │ └── / │ │ ├── forte/ │ │ └── piano/ │ └── dataset_summary.json └── synthetic-dataset/ ├── audio/ │ └── / │ └── / └── dataset_summary.json Both subsets use 12 interval classes. Intervals are named by their size (2nd, 3rd, 4th, etc.) and quality (major, minor, perfect): major_2nd — Major second (2 semitones) major_3rd — Major third (4 semitones) major_6th — Major sixth (9 semitones) major_7th — Major seventh (11 semitones) minor_2nd — Minor second (1 semitone) minor_3rd — Minor third (3 semitones) minor_6th — Minor sixth (8 semitones) minor_7th — Minor seventh (10 semitones) perfect_4th — Perfect fourth (5 semitones) perfect_5th — Perfect fifth (7 semitones) perfect_8th — Perfect octave (12 semitones) tritone — Tritone (augmented fourth/diminished fifth, 6 semitones) Statistics: Total files: 8,236 WAV files Recorded subset: 4,288 files Synthetic subset: 3,948 files Duration: 2 seconds per file Format: WAV Channels: Mono Bit depth: 16-bit

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
19 March 2026
Language
EN
Document type
Dataset