Center for Language Research
University of Aizu
   English  /  日本語     

Search this site
Ultrasound Research Methodology
As a method of observing the movements of the tongue for speech research, one-dimensional ultrasound was first used about 40 years ago (e.g. Kelsey et al., 1969, JASA 46) effectively allowing one point at a time on the tongue’s surface to be seen. Two-dimensional ultrasound has been used in speech research for 27 years (since Sonies et al., 1981, JASA 70). However, only recently have higher image quality and greater affordability made ultrasound viable for this research. Methodology used in ultrasound data collection and analysis varies widely and we are developing/testing it in our lab.
・Head Movement and Tongue Compression with a Fixed Ultrasound Probe but a Free Head
With recent growth in the number of ultrasound speech laboratories, a variety of data collection settings can be observed. In some labs, head restraints are used. These are potentially expensive, space-consuming, and may make some speakers' speech unnatural. In other labs, the head may move freely, but special equipment is used either to track head movement relative to the probe and then compensate for that movement, or to keep the head at a fixed orientation with respect to the probe. Gick, Bird and Wilson (2005) found that the simplest, most effective way to limit head movement with a reasonably high degree of accuracy is to provide a rear headrest for the subject. They also found that tongue compression was minimal for the one subject they studied. In this research, we investigate head movement relative to probe position in a cost-effective experimental set-up and we measure tongue compression in 9 subjects. A Vicon MX460 motion capture system records head and probe movement while ultrasound records tongue movement. Japanese data is provided by 9 native speakers. Subjects are seated on a stool with their heads leaning against a wall and with the ultrasound probe on a tabletop microphone stand. Stimuli consist of 60 sentences (about 5 minutes) for each of the 9 subjects.