Rhythmic structure of Hindi and English: new insights from a computational analysis.

Much information about speech rhythm is believed to be embedded in low frequency temporal modulations of the speech envelope. Using novel methods of spectral analysis we construct a spectro-temporal modulation spectrum and extract low frequency temporal modulations of spoken utterances to study the rhythmic structure of English and Hindi. The results of our spectral analysis reveal a narrower temporal bandwidth for Hindi as compared to English. We also calculate variability in syllable durations and find that variability in English is greater than Hindi. We relate temporal bandwidth of the modulation spectrum to variability in syllable duration and suggest that narrow bandwidth in the modulation spectrum implies less variability, whereas broad bandwidth implies greater variability in syllable duration. Our results also demonstrate that syllabic information is contained in low frequency temporal modulations of the speech envelope. Our results suggest that the modulation spectrum can be explored as a promising tool to study the temporal structure of language.