STUDY AND ANALYSIS OF SPEECH PARAMETERS AND BANDWIDTH COMPRESSION TECHNIQUES.
Abstract
Properties of the six stop consonants of English speech were studied with the aid of electronic gating circuitry. The apparatus made possible transient-free blanking of an adjustable-duration section at adjustable location in an input test utterance recorded on a tape loop. If desired, an excised section could be replaced with bandlimited white noise or other artificial signals. Altered and unaltered specimens were presented in random order to a listener team and responses evaluated. Results provide information on the distribution in time of recognition clues to perception of the stop consonants, and on the nature of such clues. Analog parameters representing running-average zero-crossing rate and amplitude envelope of the speech signal were extracted for 22 words spoken 10 times each by 8 speakers. These were converted to digital form and used to generate composite 20-by-20 matrices representing average plots of the one parameter vs. the other for all 80 repetitions of each word, as well as single-speaker word averages. The results bear on questions of automatic speech recognition. Bandwidth-compression experiments were also conducted, using electronic circuitry to reconstruct speech from narrowband transmitted analogs of the same two parameters. A multiplexer circuit was developed capable of switching sampling rates automatically at a desired amplitude level. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 30, 1967
- Accession Number
- AD0656478
Entities
People
- Berry O. Pyron
- Frank R. Williamson Jr.
Organizations
- Georgia Tech