A human vocal utterance corpus for perceptual and acoustic analysis of speech, singing, and intermediate vocalizations
David Gerhard · The Journal of the Acoustical Society of America · 2002
In this paper we present the collection and annotation process of a corpus of human utterance vocalizations used for speech and song research. The corpus was collected to fill a void in current research tools, since no corpus currently exists which is useful for the classification of intermediate utterances between speech and monophonic singing. Much work has been done in the domain of speech versus music discrimination, and several corpora exist which can be used for this research. A specific example is the work done by Eric Scheirer and Malcom Slaney [IEEE ICASSP, 1997, pp. 1331–1334]. The collection of the corpus is described including questionnaire design and intended and actual response characteristics, as well as the collection and annotation of pre-existing samples. The annotation of the corpus consisted of a survey tool for a subset of the corpus samples, including ratings of the clips based on a speech–song continuum, and questions on the perceptual qualities of speech and song, both generally and corresponding to particular clips in the corpus.