On transferring fundamental-frequency contours from one utterance to another

Abstract

A semiautomatic procedure is described for transferring the fundamental-frequency contour of a natural utterance to another utterance with identical semantic and syntactic attributes but with different lexical content. For sentences containing a price in cents embedded in a fixed carrier, paired comparison tests using the resynthesized original utterance as a standard indicate that listeners prefer the original pitch contours in only 60% of cases. However, the deviation from the null hypothesis of equal preference was significant at the 0.01% level.