Advances in the acoustic correlates for nasals from analysis of MRI data

MRI data for nasals is being used to simulate the nasal murmur spectrum in order to improve our understanding of nasals in speech. The data consists of American English nasals /m/, /n/, and /ng/ from one speaker [Story et al., J. Acoust. Soc. Am. 100, 537–554 (1996)]. A computer simulation model developed in our lab is being used for the purpose [Zhang and Espy-Wilson, J. Acoust. Soc. Am. (2004)]. A sufficiently good match has been obtained between the simulated and real spectra, particularly during the low frequencies (below 2000 Hz). Currently, only the maxillary and sphenoidal sinuses have been included in the model. The match between the two spectra is expected to improve further when frontal and ethomoidal sinuses are included in the model. It is our belief that this study will give us a better understanding of the acoustic manifestations of nasal manner and place in the nasal murmur spectrum and help us achieve our goal of finding speaker-independent acoustic parameters (APs) for them. It would also be interesting to see if this study can help us in finding speaker-dependent APs from the nasal murmur spectrum for our speaker recognition system. [Work supported by NSF Grant No. BCS0236707.]