Analysis of vocal tract shape and dimensions using magnetic resonance imaging: vowels.

Magnetic resonance imaging (MRI) techniques were used to gather basic data to apply in computational models of speech articulation. Two experiments were performed. In experiment 1, voice recordings from two male subjects were obtained simultaneously with axial, coronal, or midsagittal MR images of their vocal tracts while they produced the four point vowels. Area functions describing the individual tract shapes were obtained by measurements performed on the MR images. Digital filters derived from these functions were then used to resynthesize the vowel sounds which were compared, both perceptually and acoustically, with the subjects' original recordings. In experiment 2, axial images of the pharyngeal cavity were collected during the production of an ensemble of nine vowels. Plots of cross-sectional area versus the midsagittal width of the tract at different locations within the pharynx and for different vowel productions were used to derive a functional relationship between the two variables. Data from experiment 1 relating midsagittal width to cross-sectional area within the oral cavity were also examined.

Application of MRI to the analysis of speech production.

Computer models of the process of speech articulation require a detailed knowledge of the vocal tract configurations employed in speech and the application of acoustic theory to calculate the sound waveform. Almost all currently available data on vocal tract dimensions come from x-ray films and are severely limited in quantity and coherence due to restrictions on radiation dosage and intersubject differences. We are using MRI techniques to obtain the pharyngeal dimensions of speakers producing sustained vowels. The fact that MRI does not employ ionizing radiation provides speech research with the opportunity to obtain comprehensive bodies of much-needed data on the articulatory characteristics of single subjects.

Stress and vowel duration effects on syllable recognition.

Systems designed to recognize continuous speech must be able to adapt to many types of acoustic variation, including variations in stress. A speaker-dependent recognition study was conducted on a group of stressed and destressed syllables. These syllables, some containing the short vowel /I/ and others the long vowel /ae/, were excised from continuous speech and transformed into arrays of cepstral coefficients at two levels of precision. From these data, four types of template dictionaries varying in size and stress composition were formed by a time-warping procedure. Recognition performance data were gathered from listeners and from a computer recognition algorithm that also employed warping. It was found that for a significant portion of the data base, stressed and destressed versions of the same syllable are sufficiently different from one another as to justify the use of separate dictionary templates. Second, destressed syllables exhibit roughly the same acoustic variance as their stressed counterparts. Third, long vowels tend to be involved in proportionally fewer cross-vowel errors but tend to diminish the warping algorithm's ability to discriminate consonantal information. Finally, the pattern of consonant errors that listeners make as a function of vowel length shows significant differences from that produced by the computer.

Vowel height and the perception of consonant nasality.

By means of an articulatory synthesizer, the perception of the oral-nasal distinction in consonants was explored experimentally. This distinction was chosen because it is achieved by a very simple articulatory maneuver and because it is phonologically relevant in virtually every language. Lowering the velum in equal increments provided continua of CV syllables varying in size of velopharyngeal port which were divided perceptually into /d/ and /n/ categories by American English listeners. To test the hypothesis that the coarticulation of these nasal consonants with lower (more open) vowels requires a larger area of velopharyngeal coupling to give a nasal consonant precept, three oral-nasal continua incorporating the vowels /i/, /delta/, and /alpha/, respectively, were presented for identification. The results were compared with those of A. S. House and K. N. Stevens [J. Speech Hear. Disord. 21, 218-232 (1956)] and A. S. House [J. Speech Hear. Disord. 22, 190-204 (1957)] obtained with steady-state vowels and consonantal murmurs and with those of M. H. L. Hecker [J. Acoust. Soc. Am. 34, 179-188 (1962)]. Three conclusions emerged. First, the relationship between vowel height and the amount of velopharyngeal coupling needed for a nasal precept occurs in conditions where subjects are required to make linguistically relevant judgments. Second, the relationship can arise in conditions where vocalic coarticulation is present. Third, the relationship is not confined to vowels but can also be observed in the case of dynamically articulated consonants. One of the continua was also used for discrimination experiments, which yielded the classical pattern of high discriminability at the category boundary.

Reading devices for blind people.

An account is given of the social, economic, and technical factors surrounding the development of reading aids or devices for blind people. It is shown that these factors interact in such a way as to seriously constrain the approaches that can be made to device design. The progress being made in the research, development, and deployment of several types of reading devices is described to illustrate how the various constraints have affected the design goals and methods of deployment. Emphasis is laid on the development of reading machines with voice output because, ultimately, they are the most likely types of devices to be widely accepted by the blind. Looking ahead to full deployment of such devices before the end of this decade, the paper concludes with a summary of the research, development, and organizational needs that must be met.