Speech recognition

Theory and C++ implementation

Lucio Prina Ricotti, Claudio Becchetti

428 pages, parution le 26/03/1999

Ajouter à une liste

Indisponible

Résumé

Automatic speech recognition (ASR) is the technology behind the voice-triggered computer menus. Users of these systems include private telephone exchanges and real-time, on-line telephone information services. This book introduces the basic principles behind ASR programming.

Summary of contents

Introduction to ASR problems
introduction to C++
why choose C++?
foundation classes - how to use them
ASR technical details
other ASR methods
foundation classes programmers guide
speech databases - how to interface with them
speech databases
what is a speech database?
choosing the phoneme typology
commercial database description
the phoneme alphabet
quantization and sampling
I/O C++ interface to speech databases
inside the database
TIMIT characteristics
how to build a customized database
database interface programmers guide
extracting features
feature extraction problems
software description routine
technical reference of feature extraction
software routine timing
the observation engine
how the ASR "Eye" works
initialization of observation parameters
observation software
inside the observation engine
initialization of parameters
inside the implemented observation
the statistic low level engine
the HMM model
HMM training
the HMM C++ class
the HMM in literature
the HMM C++ classes - technical description
high level grammars (statistic high level engine)
grammars overview
C++ calls for grammar
technical description of grammars
technical implementation of grammars
automatic continuous recognition
automatic continuous recognition overview, modified viterbi C++ description
technical description of recognition methods
general system tests and applications
modified viterbi implementation
system enhancements and customizations
options adaptation
description of software for options
changing algorithms for customization
implementation of new algorithms