ASA 124th Meeting New Orleans 1992 October

4pMU2. Fundamental frequency tracking in the log frequency domain based on a pattern recognition algorithm.

Judith C. Brown

Media Lab., MIT, Cambridge, MA 02139

Wellesley College, Wellesley, MA 02181

Miller S. Puckette

IRCAM, Paris 75004, France

When the Fourier components of a sound with harmonic frequency components are plotted against log frequency, a constant pattern independent of fundamental frequency is obtained. A computationally efficient implementation [J. C. Brown and M. S. Puckette, ``An efficient algorithm for the calculation of a constant Q transform,'' J. Acoust. Soc. Am. (to be published 1992)] will be described. One can then calculate the cross-correlation function of these spectra with the ideal pattern, which consists of ones at the positions of the harmonic frequency components, to obtain the fundamental frequency of the input signal. Extremely accurate values of the fundamental frequency can be obtained using a method based on phase changes of the discrete Fourier transform [J. C. Brown and M. S. Puckette, ``A high resolution fundamental frequency determination based on phase changes of the Fourier transform,'' submitted to J. Acoust. Soc. Am. (1992)]. Results for a variety of musical instruments will be presented, including recent results on tracking very low frequencies produced by a cello.