Speech Spectral Moment Convergence. Voiced-Voiceless Consonant Contrasts in Whispered Speech

摘要

Continuing progress is being made in the study of the basic statistical properties of speech. Speech waveform spectral moments have been investigated, and have been found to converge in approximately 60 to 100 seconds. The result of the moment investigations are presented in part 3 of this report. An experiment in whispered speech was concluded. A word list was compiled in which pairs of words were contrasted only by a pair of consonants having voicing-non-voicing as the primary distinguishing listener que. Tests were conducted to determine the intelligibility of the consonant contrasts in the whispered mode. Investigations into the voicing patterns of speech waveforms and speech modulation by pseudo-random sequences are currently in the planning stages. (Author)