Speech Recognition: Acoustic, Phonetic and Lexical

摘要

As part of her Master's Thesis, Aull constructed a Lexical Stress Determiner for discrete words. In this document the author proposes as her thesis to make Aull's system more robust, both from a programmer's point of view and from a performance and reliability perspective. The second chapter of this thesis describes lexical stress. It explores what lexical stress is and how it might be important to a speech recognition system. The third chapter describes Aull's system for automatic detection of lexical stress in isolated words, exploring the components of her system developed by others. The fourth chapter explains the modifications that have been made Aull's system and how recognition they change system performance. The last chapter contains conclusions and some possible directions for future development.