Multi-User Real-Time Speech Processing Facility.
Abstract
Research projects requiring the VAX 11/750 and related peripherals have all focused on speech analysis, perception and recognition. We compared perceptual confusions occurring for natural and synthetic speech syllables which showed that synthetic speech is not equivalent to noisy or degraded natural speech. We conducted a study that indicated that perception of synthetic speech is improved by training. Training with fluent synthetic sentences improves performance for both isolated words and sentences. Training with isolated words improves performance on isolated words but does not improve performance on fluent synthetic sentences. A large-scale series of experiments investigated the effects of noise in a talker's ears on speech production. Words produced in noise are longer, louder and higher in pitch than words produced in the quiet. The tilt of the power spectrum decreased and formant frequencies shifted in noise. Further projects in preparation include detailed acoustic-phonetic and multi-dimensional analyses of speech produced under conditions of noise, stress and acceleration and additional research on perceptual and cognitive constraints imposed on the listener when listening to synthetic speech is continuing as well. Keywords: Speech, synthesis, analysis, perception, recognition, I/O, human factors, cognitive processes, and communication sciences.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 21, 1985
- Accession Number
- ADA164005
Entities
People
- David B. Pisoni
Organizations
- Indiana University Bloomington