Pitch-based acoustic feature analysis for the discrimination of speech and monophonic singing

David Gerhard · Canadian acoustics · 2002

this paper. To investigate the perceptual differences between talking and singing, human subjects were exposed to a corpus of singing and talking sounds, and asked first to classify each sound on a scale between speaking and singing, and then to indicate the characteristics of the sounds that lead to their judgements. The subject responses indicated that pitch is a primary factor in making this judgement. Subjects indicated many pitch-based subfeatures including vibrato (similar to Zhang's harmonic ripple), excessively low or high pitch, adherence to a musical scale, and smoothness of pitch. Other features not directly related to pitch include rhythm, rhyme, context and expectation. As will be seen later in this paper, some of these features can also be investigated using pitch as a base feature

Read the paper · More papers on PaperTik