This paper describes work on the recognition of British English intonation contours at the sentence level, using continuous density hidden Markov models (CHMMs) and the 1961 O'Connor and Arnold scheme for describing intonation contours. The scheme is presented briefly and the model optimisation procedure is described. Experimental results are presented for four tasks: transcription, type of nuclear tone, position of the voiced part of the accents and the nucleus (or principal accent). Taking into consideration the small training and test material, the results show a fairly high accuracy in all four tasks.
Bibliographic reference. Jensen, U. / Moore, Roger K. / Dalsgaard, Paul / Lindberg, Borge (1993): "Modelling of intonation contours at the sentence level using CHMMs and the 1961 o'connor and arnold scheme", In EUROSPEECH'93, 785-788.