Page 26 - Fall2013

MECHANICAL MODELS OF THE HUMAN VOCAL TRACT
Takayuki Arai
Department of Information and Communication Sciences Sophia University
Tokyo, Japan
Introduction
This article discusses the use of mechanical models of the human vocal tract in teaching students about how the vocal tract works in the formation of speech.
The idea of having a mechanical model that simulates the human vocal tract dates back at least as far as to Wolfgang von Kempelen (see Fig. 1), whose manually operated speed synthe- sizer was developed between 1769 to 1791. An account can be found at the Wikipedia site http://en.wikipedia.org /wiki/Wolfgang_von_Kempelen's_ Speaking _Machine
Highly recommended reading is the
of the Acoustical Society of America (1950) by Dudley and Tarnozy \\\[1\\\] that reviews von Kempelen’s development of his mechanical synthesizer and which also reviews the invention of analogous devices in later years. Von Kempelen summa- rized his work in a 451 page book \\\[2\\\] “Mechanismus der menschlichen Sprache bevst Bescreibung eincer sprechende Maschine (The mechanism of human speech, with a descrip- tion of a speaking machine).
In 1837, Charles Wheatstone (Fig. 2) resurrected the work of Wolfgang von Kempelen, creating an improved repli- ca of his Speaking Machine. Using new technology developed over the previous 50 years, Wheatstone was able to further analyze and synthesize components of acoustic speech, giv- ing rise to the second wave of scientific interest in phonetics. After viewing Wheatstone’s improved replica (Fig. 1) of the Speaking Machine at an exposition, a young Alexander Graham Bell set out to construct his own speaking machine with the help and encouragement of his father. Bell’s experi- ments and research ultimately led to his invention of the tele- phone in 1876, which revolutionized global communication.
Over the past few years, the present author has devel- oped and built several
physical models of the
human vocal tract.
The reader will find these described in ref- erences 3-6. These models have been found to be helpful in teaching students who are studying acoustics and speech science. Having a variety of different types of
“Having a variety of different types of vocal-tract models is important because individual models can address the different set of configurations that occur in the formation of speech.”
paper in the Journal
vocal-tract models is important because individual models can address the dif- ferent sets of configurations that occur in the formation of speech. All of the models that the author has developed demonstrate (1) the relationship between vocal-tract configuration and vowel quality, and (2) the source-filter theory of speech production \\\[5\\\]. Even with the connected-tube (CT) model \\\[2\\\], which is one of the simplest types, it is possible to demonstrate these two points. However, if one wants to teach students about the effects of tongue position and tongue movement, one
needs models where the simulated vocal tract is not straight (loosely referred to here as bent models) and one also needs adjustable models of the vocal tract. The present article focuses on these types of models.
In more recent years, a talking robot has been developed that can change the vocal tract configuration dynamically \\\[7\\\]. The author and his colleagues, for the purpose of having appropriate demonstration apparatus for students, have also developed several adjustable dynamic models of the human vocal tract for educational purposes. The models that have been developed include (1) a sliding-three-tube (S3T) model \\\[4, 10\\\], (2) Umeda and Teranishi's computer-controlled model \\\[11, 12\\\], (3) a gel-type tongue model \\\[5, 6\\\] , and (4) head-shaped models \\\[3, 12\\\]. The first two models have a straight vocal tract, and the actual area variation with length is roughly the same as for the human vocal tract. In these cases, one can demonstrate that changing the vocal-tract configuration yields different vowel qualities in real time. Therefore, learners can compare the changes in the configu- ration visually (as seen by their eyes) as well as audibly hear- ing the changes in the output sounds with their ears. These designs are nevertheless relatively simple, so the simulation of the dynamic movements with these models are less realis-
Figure 1. Replica of Von Kempelen's Speaking Machine (taken from the Wikipdia site)
Figure 2. Charles Wheatstone
tic than for the other two models.
Bent vocal-tract models are suitable in demonstrations of how one produces vowels with the use of one’s speech organs. A common question that learners often ask is why the vocal tract isn’t placed in our heads. Static bent
Mechanical Models of Vocal Tract 25