Contributed by:
	De Lathauwer Lieven
	K.U.Leuven, E.E. Dept.- ESAT
	K. Mercierlaan 94
B-3001 Leuven (Heverlee)
Belgium
	Lieven.DeLathauwer@esat.kuleuven.ac.be
Description:
	The dataset is a real-valued (5x10x13)-array; basically, it was obtained as
	follows. High-quality audio recordings and cine-fluorograms were made of
	five English-speaking test persons while saying sentences of the form ``Say
	h({\em vowel})d again'' (substitution: ``heed, hid, hayed, head, had, hod,
	hawed, hoed, hood, who'd''). For each of these 10 vowels an acoustic
	reference moment was defined and the corresponding 5 (corresponding to the
	different speakers) frames in the film located. Next, speaker-dependent
	reference grids, taking into account the anatomy of each test person, were
	defined and superimposed on the remaining x-ray images. The grids consisted
	of 13 equidistant lines, in the region from epiglottis to tongue tip, more
	or less perpendicular to the midline of the vocal tract. The array entries
	now consist of the distance along the grid lines between the surface of the
	tongue and the harder upper surface of the vocal tract. The values are given
	in centimeters and have been measured to the nearest 0.5 mm. For a more
	extensive description of the experiment, we refer to [1].
Sampling:
Number:
Inputs:
Outputs:
	x: speakers
	y: vowels
	z: positions
References:
	[1] R. Harshman, P. Ladefoged, L. Goldstein,
	``Factor Analysis of Tongue Shapes'',
	J. Acoust. Soc. Am., Vol. 62, No. 3, Sept. 1977, pp. 693-707.
	
	[2] L. De Lathauwer, Signal Processing based on Multilinear Algebra,
	Ph.D. Thesis, K.U.Leuven, E.E. Dept., Sept. 1997.
Properties:
Columns:
	Column 1: (speaker number - 1) x 10 + vowel number
	Columns 2-14: displacement values
Category:
	Biomedical systems