VtoV: a perceptual cue for rhythm identification

Current metrics for the quantification of speech rhythm take into account parameters not easily detectable by listeners. To overcome this limit, in this study we propose a new model based on a parameter that account for listeners' ability to discriminate between different rhythmic patterns. Starting from the results of a spectro-acoustic analysis conducted on singing, we found that Perceptual Centres align close to Vowel Onset Points (VOP). To test the perceptual relevance of interval between two consecutive VOPs, that we call VtoV intervals, we analyzed a multilingual corpus of TV news, advertisings and recited speech. The signal was segmented into vocalic/consonantal portions and into VtoV intervals. The standard deviation of all parameters was calculated. The results of the analysis show that VtoV is a crucial parameter both to classify languages on a rhythmic basis and to account for intra-linguistic speech style variations.