Second International Conference on Spoken Language Processing (ICSLP'92)

Banff, Alberta, Canada
October 13-16, 1992

Analysis of Phonation Type Using Laryngographic Techniques

John H. Esling (1,2), B. Craig Dickson (1,2), Roy C. Snell (2)

(1) Department of Linguistics, University of Victoria, Victoria, B.C., Canada
(2) Speech Technology Research, Ltd., Victoria, B.C., Canada

In order to develop a laryngographic (Lx) waveform measurement algorithm for phonation types which is independent of frequency variation, Lx and acoustic signals of a sustained vowel at seven pitch increments for a set of standard models of laryngeal voice qualities are captured and analyzed in the Computerized Speech Lab (CSL) environment. Auditory assessments are used to categorize the data; a procedure is applied to eliminate DC float in the Lx signal and flatten the baseline; a pitch-extraction algorithm is introduced to compute pitch and jitter; speed-quotient and open-quotient techniques are applied to derive Lx-period ratio comparisons; and a cepstral procedure is used to establish periodicity indices for the pitch-differentiated waveform samples. The research objective is to reliably identify and distinguish degrees of breathy voice, whispery voice, harsh voice or ventricular voice, where the creaky voice/ modal voice/ falsetto frequency dimension is viewed as a confounding factor in the recognition of phonation types in the breathiness/ harshness dimension. The practical objective is to develop speech and language technology with potential clinical applications in the analysis of pathological phonatory qualities.

Full Paper

Bibliographic reference.  Esling, John H. / Dickson, B. Craig / Snell, Roy C. (1992): "Analysis of phonation type using laryngographic techniques", In ICSLP-1992, 1107-1110.