Most of the speech analysis methods developed up to date have been based on the autocorrelation function or power spectrum, i. e., the second order statistics of the signa!. In this paper it is shown that higher order statistics (cumulants) are also very useful in the study and characterization of the speech. A new pitch determination algorithm based on third order statistics is presented and tested in noise corrupted speech, at different levels of signa! to noise ratio, and with different kinds of noise. The results show that this new algorithm give a much better estimation of the pitch that the conventional autocorrelation method in al! the cases.

Peer Reviewed