Page 56 - 2016Winter
P. 56
Speech Production Research
traveling through it. These two (or more) velocities cannot be captured by an electrical analogue. As a result, the de- tails of various phenomena are incorrectly modeled, includ- ing the meaning of the DC flow computed by inverse filter- ing (see Shadle et al., 1999) and the mechanism by which the noise source is modulated by the voice source in voiced fricatives (Jackson and Shadle, 2000). More importantly, the geometric details incorporated in such a synthesizer make one think it is more physically accurate than it is.
Computational fluid dynamics (CFD) allows the fluid prop- erties to be modeled. Although CFD has been a very use- ful third domain in which to study phonation (Zhang et al., 2002; Zhao et al., 2002) and, to some extent, turbulence noise in fricatives (Adachi and Honda, 2003), it is not at all straightforward to extract the sound that results from a given flow field. The addition of turbulence imposes a big computational burden. Although this method combines aerodynamics and acoustics successfully, it relies on com- puter power to handle the fine details of fluid motion rather than simplifying them so that only the aspects of the flow known to be crucial to the sound produced are modeled. It thus offers a more physically realistic method of predicting the flow field and the sound produced but not yet a different way to think about that process.
Some smaller studies offer a way forward. For instance, by relaxing some of the assumptions in circuit analogues, it was possible to test how much of a difference each makes (Davies et al., 1993). Particular articulatory-aerodynamic interac- tions have been studied, such as whether intraoral pressure has an appreciable effect on tongue motion (Mooshammer et al., 1995) or whether articulatory changes such as cavity expansion have an appreciable effect on phonation (West- bury, 1983). The interaction of a sound wave with a cloud of turbulence has been studied to explain the production of [s] (Howe and McGowan, 2005). Measurements in a mechani- cal model of the vocal folds and tract were interpreted in terms of a combined flow and acoustic field (Barney et al., 1999). It has long been accepted that vocal fold vibration is a complex system. To predict the effect of any physical change accurately, a vocal fold model must include aerodynamic, mechanical, and acoustic elements. We have not yet arrived at comparable models of supraglottal aeroacoustic sources.
Final Thoughts
The study of speech production is an ongoing endeavor. The well-accepted wisdom of any era is subject to revision in the future with the advent of new ideas, new instrumentation, and new research. This cycle of repudiation and revision will surely happen again with the ideas we hold dear today. These seven myths were once among the most prominent theories of their day, and some aspects of them are still open to de- bate. We have presented them to describe the history and development of the field.
Sometimes the major theories of the past were converted to “myth” status by the development of instruments and methodologies that provided additional perspective. Such changes in perspective occurred when the sound spectro- graph revealed the true nature of speech sound sequencing and when sophisticated imaging techniques revealed what the 3-D vocal tract looked like in motion. In other cases, existing instruments or slight modifications added the in- formation needed to rethink our idea, such as when studies included frequencies higher than telephone bandwidth and when bite block studies revealed the variety of ways one can produce the same speech sound. Most importantly, open- ness of thought and discussion of ideas allowed us to change our theories and models, even when the models were more elegant and more appealing than the true data.
Acknowledgments
We would like to thank Ereni Sevasti, the subject for Figure 1, and Richard Lissemore, who recorded her speech for us.
Biosketches
Maureen Stone is a professor at the Uni- versity of Maryland School of Dentistry, Baltimore, with a joint appointment in the Department of Neural and Pain Sciences and the Department of Ortho- dontics. She is also director of the Vocal Tract Visualization Laboratory. She is a
speech scientist by profession and has spent her career using imaging techniques such as MRI and ultrasound to study the behavior of the human tongue during speech. She began her career as a research scientist at the National Institutes of Health and then at Johns Hopkins University before coming to the University of Maryland. She is a Fellow of the Acous- tical Society of America.
54 | Acoustics Today | Winter 2016