Paul Boersma points out in his article 'ACCURATE SHORT-TERM ANALYSIS
OF THE FUNDAMENTAL FREQUENCY AND THE HARMONICS-TO-NOISE RATIO OF A
SAMPLED SOUND'
(http://www.fon.hum.uva.nl/paul/papers/Proceedings_1993.pdf) that
there are two ways to improve the use of an autocorrelation method
to track fundamental frequency in speech. One (correcting the
autocorrelation for the effects of the window function) is simple,
but the interpolations he suggests for best estimation of the lag
and the height of that peak (especially the sin(x)/x one) is not!
Has anyone implemented these procedures in Matlab by any chance, and
are willing to share their code? Thanks! Yours - Stuart --
Stuart Rosen, PhD Professor of Speech and Hearing Science UCL Speech, Hearing and Phonetic Sciences 2 Wakefield Street London WC1N 1PF, England Telephone numbers:
Home page: http://www.phon.ucl.ac.uk/home/stuart Request an inspection copy of Signals and Systems for Speech and Hearing, 2nd edition by Stuart Rosen & Peter Howell http://info.emeraldinsight.com/promo/signals.htm |