Abstract
This study proposes an automatic method for the detection of pauses and identification of pause types in conversational speech for the purpose of measuring the effects of Friedreich's Ataxia (FRDA) on speech. Speech samples of ∼ 3 minutes were recorded from 13 speakers with FRDA and 18 healthy controls. Pauses were measured from the intensity contour and fit with bimodal lognormal distributions using the Expectation-Maximization algorithm in Matlab©. In the speakers with FRDA, both modes in the pause distributions had significantly larger means, with disproportionately fewer pauses associated with the first mode. From this preliminary study, it is concluded that distributional analysis of pause duration holds promise as a useful method of measuring the effects of FRDA on functional speech.
Keywords:
Acknowledgements
The authors thank the participants involved in this research. This project was funded by a research grant from the Friedreich's Ataxia Research Alliance (FARA), USA, and the Friedreich's Ataxia Research Association, Australasia.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
Notes
1. The anti-alias filter and relatively high sampling rate theoretically is not necessary for pause analysis, but was applied to the files for future spectral studies.
2. Although the proposed measurement and analysis of pause duration is automatic, it should be noted that the preparation of the samples does require man hours. For this study, digitizing and saving the recordings required about 7 minutes per speaker. The listening and editing to the 3 minute samples required an additional average of 6 minutes per speaker, but is highly dependent on the experience of the editor.
3. Another example of a single factor responsible for bimodality in a complex phenomenon is gender in a distribution of human height (Joiner, Citation1975; but see Schilling, Watkins, and Watkins, Citation2002).
4. The standard deviation of log duration is akin to coefficient of variation, and can be a useful index of variation for phenomena in which increases in standard deviation coincide with increases in mean, such as that occur in speech segment duration (see Rosen, Citation2005).