Abstract
Articulatory data have been collected by a computer-controlled microbeam system. The locations of lead pellets placed on a speaker's articulators are recorded as x, y coordinates along with the acoustic signal and a train of timing pulses which synchronize the frames of X-ray data with the acoustic signal. Some 830 utterances have been recorded to date, resulting in a database of nearly 120 million bytes. We describe the graphic techniques by which those data are examined; namely, a simultaneous display of articulatory features such as velum lowering, lip and jaw closure, and tongue motion, together with the spectral information of the corresponding speech. These displays are also annotated with phonetic transcriptions obtained automatically by pattern-matching techniques. Acoustic representations of speech signals, in a simplified spectrographic format, are contrasted with articulatory measures, and implications are discussed concerning the difficulty of the automatic recognition of speech via conventional input processing.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call