Simple Prosodic Peak Estimation under Segmental Perturbations

Despite the apparent simplicity, measuring the position of f0 peaks can produce unexpected results. In any model of intonation where the fundamental frequency is derived from both an intonation component and a segmental component (e.g. consonantal perturbations or the intrinsic f0 shifts of vowels), the measured peak position does not equal the peak of the intonation component. The difference can sometimes be unexpectedly large. Indeed, measured peak positions can falsely imply a phonological distinction in the intonation where none really exists. A modified measurement technique is presented that is less sensitive to segmental effects than directly measuring the f0 maximum. It is tested mathematically in a superposition model of intonation and also on speech data.