Speech Recognition: Acoustic-Phonetic Knowledge Acquisition and Representation.
Abstract
Our long-term research goal is the development and implementation of speaker-independent continuous speech recognition systems. It is our conviction that the proper utilization of speech-specific knowledge is essential for such advanced systems. Our research is thus directed toward the acquisition of acoustic-phonetic and lexical knowledge, and the application of this knowledge to speech recognition algorithms. We continued our investigation into the contextual variations of speech sounds, emphasizing on the role of the syllable in these variations. Our analysis revealed that the acoustic realization of a stop depends greatly on its position within a syllable. In order to represent and utilize this information in speech recognition, we have adopted a hierarchical syllable description that enables us to specify the constraints in terms of an immediate constituent grammar. We will continue to quantify the effect of context on the acoustic realization of phonemes using larger constituent units such as syllables. In addition, we will develop a grammar to describe the relationship between phonemes and acoustic segments, and a parser that will make use of such a grammar for phonetic recognition and lexical access.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 24, 1987
- Accession Number
- ADA187293
Entities
People
- Victor W. Zue
Organizations
- Massachusetts Institute of Technology