AVP-LVT Vocal Percussion Dataset
<p>The <strong>AVP-LVT dataset</strong> contains vocal percussion utterances from two publicly available datasets: the personal subset of the <a href="https://dl.acm.org/doi/abs/10.1145/3356590.3356844">Amateur Vocal Percussion (AVP) dataset</a> and the third subset of the <a href="https://repositorio-aberto.up.pt/handle/10216/105309">Live Vocalised Transcription (LVT) dataset</a>. They contain vocal percussion sounds from a total of 48 participants with little or no experience in beatboxing.</p> <p>The <strong>AVP dataset</strong> contains a total of 4873 vocal percussion sound events recorded by 28 participants and with four annotated labels: kick drum, snare drum, closed hi-hat, and opened hi-hat. For each participant, four files contain repetitions of vocal percussion sounds of the same class and one file corresponds to a freestyle improvisation with these. The <strong>LVT dataset</strong> contains a total of 841 vocal percussion sound events recorded by 20 participants with three annotated labels: kick drum, snare drum, and closed hi-hat. For each participant, one file contains a predictable beatbox-style phrase repeated four times and another file contained a freestyle improvisation with the sounds used in the phrase file.</p> <p>This dataset expands the original annotations of both datasets (onsets and instrument labels) so as to include <em>syllabic annotations</em> of vocal percussion sounds. These annotations are composed of a first <em>onset phoneme</em>, usually plosive or fricative, and a second <em>coda phoeneme</em>, usually a vowel, a breath sound ("h"), or silence ("x"). These phonemes were annotated following notation conventions from the International Phonetic Alphabet (IPA).</p> <p>The files included here are the recordings and the annotations of the AVP dataset and the annotations of the LVT dataset. To incorporate the audio files from the LVT dataset, please follow the steps outlined in the file entitled "Instructions_to_build_AVP-LVT_Dataset.rtf". This file also contains information about the train-evaluation split for research purposes.</p>
ShareScore
36/100
Overall dataset sharing score
Score breakdown
These five areas show where the dataset supports — or may limit — practical reuse.
- Stewardship
- 4
- Harmonization
- 4
- Access
- 16
- Reuse readiness
- 8
- Engagement
- 4