Amateur Vocal Percussion Dataset

Dataset of vocal imitations of percussion sounds. It includes a total of 9780 utterances in 280 audio files recorded by 28 participants with little or no experience in beatboxing. Onsets and labels for kick drum, snare drum, closed hi-hat, and opened hi-hat are annotated in separate .csv files. All audio files were recorded with a MacBook Pro's built-in mic in an indoor environment and they are distributed into two modalities: personal imitations and fixed imitations. For each modality and participant, four files contain repetitions of utterances of the same class and one file corresponds to a rhythmic improvisation with these. <strong>Update from V3</strong>:<em> Onset and coda phoneme labels added, sample noise audio file added.</em>