Low-frequency Fourier Analysis of Speech Rhythm
Sam Tilsen, Keith Johnson · UC Berkeley Phonology Lab Annual Reports · 2007
We present a new methodology for studying speech rhythm, based upon low-frequency Fourier analysis of the amplitude envelope of bandpass-filtered speech.Rather than quantifying rhythm with time-domain measurements of interval durations, we use frequency-domain representations to characterize speech rhythm with a metric based on a rhythm spectrum.In this paper we describe our method in detail, and using the Buckeye corpus, we illustrate and discuss approaches to characterizing rhythm with low-frequency spectral information.