Accession Number : ADA299988

Title :   Text-To-Speech Phrasing Enhancement System Using Neural Networks.

Descriptive Note : Professional paper,

Corporate Author : NAVAL COMMAND CONTROL AND OCEAN SURVEILLANCE CENTER RDT AND E DIV SAN DIEGO CA

Personal Author(s) : Julig, Louise F.

PDF Url : ADA299988

Report Date : AUG 1995

Pagination or Media Count : 7

Abstract : Much progress has been made in computer text-to-speech systems over the past several years. In particular, the Macintosh computer systems now provide the PlainTalk Text-To-Speech synthesizer which is capable of using high quality voices with various attributes to convert text to synthesized speech. Though these new voices and speech synthesizer are great improvements over the previous Macintalk system the synthesized speech still sounds far from natural. One attribute which could add greatly to the naturalness of the speech is improved phrasing. The PlainTalk Text-To-Speech synthesizer provides the means to embed speech commands within text to modify the spoken output. The purpose of this project is to build a neural network which through supervised learning will produce an algorithm for embedding pitch controls in text which will produce more natural sounding emphasis variations for the spoken output.

Descriptors :   *NEURAL NETS, *SPEECH ANALYSIS, ALGORITHMS, SOFTWARE ENGINEERING, REPRINTS, DISTRIBUTED DATA PROCESSING, COMPUTER PROGRAMMING, LEARNING MACHINES, INPUT OUTPUT PROCESSING, RULE BASED SYSTEMS, PARALLEL PROCESSING, WORDS(LANGUAGE), SPEECH RECOGNITION, PATTERN RECOGNITION, COMPUTATIONAL LINGUISTICS, SOUND PITCH, PHRASE STRUCTURE GRAMMARS, CONTEXT SENSITIVE GRAMMARS, PITCH DISCRIMINATION, PHONEMES.

Subject Categories : Cybernetics
      Linguistics

Distribution Statement : APPROVED FOR PUBLIC RELEASE