Index

1.4 Baum-Welch Re-Estimation, 8.5 Embedded Training using HEREST

accumulators

accuracy figure

3.4.1 Step 11 - Recognising the Test Data

ACCWINDOW

ADDDITHER

5.8.11 ALIEN and NOHEAD File Formats

ALIEN

all-pole filter

ALLOWCXTEXP

3.4.1 Step 11 - Recognising the Test Data, 11.8 Word Network Expansion

ALLOWXWRDEXP

analysis

FFT-based: 3.1.5 Step 5 - Coding the Data
LPC-based: 3.1.5 Step 5 - Coding the Data

ANON

5.1 General Mechanism

AT command

3.2.2 Step 7 - Fixing the Silence Models, 9.7 Miscellaneous Operations

AU command

3.3.2 Step 10 - Making Tied-State Triphones, 3.3.2 Step 10 - Making Tied-State Triphones, 3.5 Running the Recogniser Live, 9.5 Tree-Based Clustering

audio output

5.9 Direct Audio Input/Output, 5.9 Direct Audio Input/Output

audio source

AUDIOSIG

average log probability

11.4 Bigram Language Models

back-off bigrams

ARPA MIT-LL format: 11.4 Bigram Language Models

backward probability

Baum-Welch algorithm

1.4 Baum-Welch Re-Estimation, 1.4 Baum-Welch Re-Estimation

Baum-Welch re-estimation

embedded unit: 8.5 Embedded Training using HEREST
isolated unit: 8.4 Isolated Unit Re-Estimation using HREST

Bayes' Rule

1.2 Isolated Word Recognition

beam width

8.5 Embedded Training using HEREST, 12.1 Decoder Operation

<BeginHMM>

7.8 Binary Storage Format, 9.2 Constructing Context-Dependent Models

binary chop

10.4 Parameter Smoothing

binary storage

binning

Boolean values

1.6 Continuous Speech Recognition, 2.3.2 Training Tools, 8.1 Training Strategies

bootstrapping

byte swapping

4.9 Byte-swapping of HTK data files, 5.2 Speech Signal Processing

byte-order

BYTEORDER

C-heaps

5.3 Linear Prediction Analysis, 5.4 Filterbank Analysis

CEPLIFTER

cepstral analysis

filter bank: 5.4 Filterbank Analysis
liftering coefficient: 5.3 Linear Prediction Analysis
LPC based: 5.3 Linear Prediction Analysis
power vs magnitude: 5.4 Filterbank Analysis

cepstral coefficients

liftering: 5.3 Linear Prediction Analysis

cepstral mean normalisation

CFWORDBOUNDARY

CH command

check sums

CHKHMMDEFS

Choleski decomposition

3.3.1 Step 9 - Making Triphones from Monophones, 9.2 Constructing Context-Dependent Models

CL command

cloning

3.3.1 Step 9 - Making Triphones from Monophones, 3.3.1 Step 9 - Making Triphones from Monophones, 9.2 Constructing Context-Dependent Models

cluster merging

3.3.2 Step 10 - Making Tied-State Triphones, 9.4 Data-Driven Clustering

clustering

data-driven: 9.4 Data-Driven Clustering
tracing in: 9.5 Tree-Based Clustering
tree-based: 9.5 Tree-Based Clustering

CO command

codebook

7.6 Discrete Probability HMMs

codebook exponent

codebooks

3.1.5 Step 5 - Coding the Data

coding

command line

arguments: 3.1.5 Step 5 - Coding the Data, 4.1 The Command Line
arguments: 3.1.5 Step 5 - Coding the Data, 4.1 The Command Line
ellipsed arguments: 4.2 Script Files
integer argument formats: 4.1 The Command Line
options: 2.2 Generic Properties of a HTK Tool, 4.1 The Command Line
options: 2.2 Generic Properties of a HTK Tool, 4.1 The Command Line
script files: 3.1.5 Step 5 - Coding the Data

command line options

compression

configuration files

2.2 Generic Properties of a HTK Tool, (, )

default: 4.3 Configuration Files
display: 4.4 Standard Options
format: 4.3 Configuration Files
types: 4.3 Configuration Files

configuration parameters

operating;tex2html_html_special_mark_quot;environment: 4.10 Summary
switching: 8.6 Single-Pass Retraining

configuration variables

display of: 4.3 Configuration Files
summary: 14 Configuration Variables

confusion matrix

9.2 Constructing Context-Dependent Models

context dependent models

continuous speech recognition

covariance matrix

cross-word network expansion

9.2 Constructing Context-Dependent Models

cross-word triphones

data insufficiency

2.3.1 Data Preparation Tools, 3.1 Data Preparation

data preparation

DC command

6.4 Editing Label Files, 11.7 Constructing a Dictionary

DE command

decision tree-based clustering

decision trees

loading and storing: 9.5 Tree-Based Clustering

decoder

12 Decoding

alignment mode: 12.2 Decoder Organisation
evaluation: 12.3 Recognition using Test Databases
forced alignment: 12.5 Generating Forced Alignments
live input: 12.6 Recognition using Direct Audio Input
N-best: 12.7 N-Best Lists and Lattices
operation: 12.1 Decoder Operation
organisation: 12.2 Decoder Organisation
output formatting: 12.5 Generating Forced Alignments
output MLF: 12.3 Recognition using Test Databases
progress reporting: 12.3 Recognition using Test Databases
recognition mode: 12.2 Decoder Organisation
rescoring mode: 12.2 Decoder Organisation
results analysis: 12.4 Evaluating Recognition Results
trace output: 12.3 Recognition using Test Databases

decompression;tex2html_html_special_mark_quot;filter

8.4 Isolated Unit Re-Estimation using HREST

defunct mixture components

defunct mixtures

deleted interpolation

10.4 Parameter Smoothing

deletion errors

delta coefficients

DELTAWINDOW

11 NetworksDictionaries and Language Models

dictionaries

dictionary

construction: 3.1.2 Step 2 - the Dictionary, 11.7 Constructing a Dictionary
construction: 3.1.2 Step 2 - the Dictionary, 11.7 Constructing a Dictionary
edit commands: 11.7 Constructing a Dictionary
entry: 3.1.2 Step 2 - the Dictionary
format: 3.1.2 Step 2 - the Dictionary
formats: 11.7 Constructing a Dictionary
output symbols: 11.7 Constructing a Dictionary

digit recogniser

direct audio input

11.4 Bigram Language Models

DISCOUNT

DISCRETE

7.6 Discrete Probability HMMs, 10 Discrete and Tied-Mixture Models

discrete HMM

output probability scaling: 7.6 Discrete Probability HMMs

discrete HMMs

discrete probability

7.1 The HMM Parameters, 7.6 Discrete Probability HMMs

discrete;tex2html_html_special_mark_quot;data

10.1 Modelling Discrete Sequences

DISCRETEHS

DP command

duration parameters

7.9 The HMM Definition Language

duration vector

EBNF

2.3.3 Recognition Tools, 11.3 Building a Word Network with HPARSE

edit commands

single letter: 6.4 Editing Label Files

edit file

3.2.1 Step 6 - Creating Flat Start Monophones

embedded re-estimation

embedded training

1.6 Continuous Speech Recognition, 2.3.2 Training Tools, 8.1 Training Strategies, 8.5 Embedded Training using HEREST

<EndHMM>

5.10 Multiple Input Streams

energy suppression

ENORMALISE

4.3 Configuration Files, 5.5 Energy Measures

environment variables

4.10 Summary

error message

format: 15 Error and Warning Codes

error number

structure of: 15 Error and Warning Codes

error numbers

structure of: 4.5 Error Reporting

errors

full listing: 15 Error and Warning Codes

ESCALE

Esignal

ESPS files

6.2.4 SCRIBE Label Files, 12.5 Generating Forced Alignments

EX command

extended Backus-Naur Form

2.3.4 Analysis Tool, 12.4 Evaluating Recognition Results

extensions

mfc: 3.1.5 Step 5 - Coding the Data
scp: 3.1.5 Step 5 - Coding the Data
wav: 3.1.3 Step 3 - Recording the Data

Figure of Merit

file formats

ALIEN: 5.8.11 ALIEN and NOHEAD File Formats
Audio Interchange (AIFF): 5.8.7 AIFF File Format
Esignal: 5.7.2 Esignal Format Parameter Files, 5.8.2 Esignal File Format
Esignal: 5.7.2 Esignal Format Parameter Files, 5.8.2 Esignal File Format
HTK: 5.8.1 HTK File Format
NIST: 5.8.4 NIST File Format
NOHEAD: 5.8.11 ALIEN and NOHEAD File Formats
OGI: 5.8.9 OGI File Format
SCRIBE: 5.8.5 SCRIBE File Format
Sound Designer(SDES1): 5.8.6 SDES1 File Format
Sun audio (SUNAU8): 5.8.8 SUNAU8 File Format
TIMIT: 5.8.3 TIMIT File Format
WAVE: 5.8.10 WAVE File Format

file;tex2html_html_special_mark_quot;formats

HTK: 5.7.1 HTK Format Parameter Files

files

adding checksums: 5.13 Copying and Coding using HCOPY
compressing: 5.13 Copying and Coding using HCOPY
configuration: 4.3 Configuration Files
copying: 5.13 Copying and Coding using HCOPY
listing contents: 5.12 Viewing Speech with HLIST
network problems: 4.8 Input/Output via Pipes and Networks
opening: 4.8 Input/Output via Pipes and Networks
script: 4.2 Script Files
VQ codebook: 5.11 Vector Quantisation

filters

8.3 Flat Starting with HCOMPV

fixed-variance

flat start

2.3.2 Training Tools, 3.1.4 Step 4 - Creating the Transcription Files, 3.2.1 Step 6 - Creating Flat Start Monophones, 8.1 Training Strategies, 8.3 Flat Starting with HCOMPV

float values

2.3.4 Analysis Tool, 12.4 Evaluating Recognition Results

FOM

FORCECXTEXP

3.4.1 Step 11 - Recognising the Test Data, 11.8 Word Network Expansion

forced alignement

12.5 Generating Forced Alignments

forced alignment

3.2 Creating Monophone HMMs

forced recognition

FORCELEFTBI

FORCEOUT

FORCERIGHTBI

forward probability

forward-backward

embedded: 8.5 Embedded Training using HEREST
isolated unit: 8.4 Isolated Unit Re-Estimation using HREST

Forward-Backward algorithm

full rank covariance

Gaussian mixture

Gaussian pre-selection

generalised triphones

global.ded

7.9 The HMM Definition Language

global options

global options macro

global speech variance

grammar

3.4.1 Step 11 - Recognising the Test Data

grammar scale factor

grand variance

9.3 Parameter Tying and Item Lists

HALIGN(1.5)

Hamming Window

2.1 HTK Software Architecture, 5.9 Direct Audio Input/Output

HAUDIO

HBUILD

2.3.3 Recognition Tools, 11.4 Bigram Language Models, 11.5 Building a Word Network with HBUILD, (, )

HCODE(1.5)

2.3.2 Training Tools, 3.2.1 Step 6 - Creating Flat Start Monophones, 8.1 Training Strategies, 8.3 Flat Starting with HCOMPV, (, )

HCOMPV

HCONFIG

2.3.1 Data Preparation Tools, 3.1.5 Step 5 - Coding the Data, 4.2 Script Files, 5.13 Copying and Coding using HCOPY, 10.2 Using Discrete Models with Speech, (, )

HCOPY

HDICT

5.8.11 ALIEN and NOHEAD File Formats

HDMAN

2.3.3 Recognition Tools, 3.1.2 Step 2 - the Dictionary, 11.7 Constructing a Dictionary, (, )

HEADERSIZE

HEREST

1.6 Continuous Speech Recognition, 2.3.2 Training Tools, 3.2.1 Step 6 - Creating Flat Start Monophones, 8.1 Training Strategies, 8.5 Embedded Training using HEREST, 9.1 Using HHED, (, )

HGRAF

2.3.2 Training Tools, 3.2.2 Step 7 - Fixing the Silence Models, 3.5 Running the Recogniser Live, 8.1 Training Strategies, 9 HMM System Refinement, (, )

HHED

HINIT

1.4 Baum-Welch Re-Estimation, 1.4 Baum-Welch Re-Estimation, 2.3.2 Training Tools, 4.2 Script Files, 8.1 Training Strategies, (, )

HIPASS

10.3 Tied Mixture Systems

HK command

HLAB2NET(1.5)

HLABEL

2.3.1 Data Preparation Tools, 3.1.4 Step 4 - Creating the Transcription Files, 3.3.1 Step 9 - Making Triphones from Monophones, 6.4 Editing Label Files, 12.5 Generating Forced Alignments, (, )

HLED

HLIST

2.3.1 Data Preparation Tools, 5.12 Viewing Speech with HLIST, (, )

HLM

2.1 HTK Software Architecture, 11.4 Bigram Language Models

HLSTATS

2.3.1 Data Preparation Tools, 11.4 Bigram Language Models, (, )

HMATH

2.1 HTK Software Architecture, 4 The Operating Environment

HMEM

2.1 HTK Software Architecture, 4 The Operating Environment, 4.7 Memory Management

HMM

binary storage: 3.3.1 Step 9 - Making Triphones from Monophones
build philosphy: 2.3.2 Training Tools
cloning: 3.3.1 Step 9 - Making Triphones from Monophones
definition files: 3.2.1 Step 6 - Creating Flat Start Monophones
definitions: 1.2 Isolated Word Recognition, 7 HMM Definition Files
definitions: 1.2 Isolated Word Recognition, 7 HMM Definition Files
editor: 2.3.2 Training Tools
instance of: 1.6 Continuous Speech Recognition
parameters: 7.1 The HMM Parameters
triphones: 3.3 Creating Tied-State Triphones

HMM definition

stream weight: 7.2 Basic HMM Definitions
basic form: 7.2 Basic HMM Definitions
binary storage: 7.8 Binary Storage Format
covariance matrix: 7.2 Basic HMM Definitions
formal syntax: 7.9 The HMM Definition Language
global features: 7.2 Basic HMM Definitions
global options: 7.9 The HMM Definition Language
global options macro: 7.2 Basic HMM Definitions
macro types: 7.3 Macro Definitions
macros: 7.3 Macro Definitions
mean vector: 7.2 Basic HMM Definitions
mixture components: 7.2 Basic HMM Definitions
multiple data streams: 7.2 Basic HMM Definitions
stream weight: 7.2 Basic HMM Definitions
symbols in: 7.2 Basic HMM Definitions
tied-mixture: 7.5 Tied-Mixture Systems
transition matrix: 7.2 Basic HMM Definitions

HMM lists

6.4 Editing Label Files, 7.4 HMM Sets, 7.4 HMM Sets, 8.5 Embedded Training using HEREST

HMM name

HMM refinement

9 HMM System Refinement

HMM sets

types: 7.4 HMM Sets

HMM tying

HMODEL

1.6 Continuous Speech Recognition, 2.1 HTK Software Architecture

HNET

HParm

SILENERGY: 5.9 Direct Audio Input/Output
SILGLCHCOUNT: 5.9 Direct Audio Input/Output
SILMARGIN: 5.9 Direct Audio Input/Output
SILSEQCOUNT: 5.9 Direct Audio Input/Output
SPCGLCHCOUNT: 5.9 Direct Audio Input/Output
SPCSEQCOUNT: 5.9 Direct Audio Input/Output
SPEECHTHRESH: 5.9 Direct Audio Input/Output

HPARSE

2.3.3 Recognition Tools, 3.1.1 Step 1 - the Task Grammar, 11.3 Building a Word Network with HPARSE, (, )

HParse format

1.6 Continuous Speech Recognition, 2.1 HTK Software Architecture, 12.2 Decoder Organisation, 12.2 Decoder Organisation

compatibility mode: 11.3 Building a Word Network with HPARSE
in V1.5: 11.3 Building a Word Network with HPARSE
variables: 11.3 Building a Word Network with HPARSE

HQUANT

2.3.1 Data Preparation Tools, (, )

HREC

HREST

1.4 Baum-Welch Re-Estimation, 2.3.2 Training Tools, 8.1 Training Strategies, (, )

HRESULTS

2.3.4 Analysis Tool, 12.4 Evaluating Recognition Results, (, )

HSGEN

2.3.3 Recognition Tools, 3.1.3 Step 3 - Recording the Data, 11.6 Testing a Word Network using HSGEN, (, )

HSHELL

2.1 HTK Software Architecture, 4 The Operating Environment

HSIGP

HSKind

HSLAB

2.3.1 Data Preparation Tools, 3.1.3 Step 3 - Recording the Data, (, )

HSMOOTH

2.3.2 Training Tools, 10.4 Parameter Smoothing, (, )

HSPIO(1.5)

HTKResearch(V1.5)

HTRAIN

HUTIL

1.5 Recognition and Viterbi Decoding, 1.6 Continuous Speech Recognition, 2.3.3 Recognition Tools, 3.2.3 Step 8 - Realigning the Training Data, 3.4.1 Step 11 - Recognising the Test Data, 12.2 Decoder Organisation, (, )

HVITE

HVQ

HWAVE

HWAVEFILTER

5.8.4 NIST File Format

insertion errors

integer values

<InvCovar>

isolated word training

3.3.1 Step 9 - Making Triphones from Monophones, 9.3 Parameter Tying and Item Lists

item lists

indexing: 9.3 Parameter Tying and Item Lists
pattern matching: 9.3 Parameter Tying and Item Lists

JO command

10.3 Tied Mixture Systems

K-means clustering

6 Transcriptions and Label Files

label files

ESPS format: 6.2.2 ESPS Label Files
HTK format: 6.2.1 HTK Label Files
SCRIBE format: 6.2.4 SCRIBE Label Files
TIMIT format: 6.2.3 TIMIT Label Files

labels

changing: 6.4 Editing Label Files
context dependent: 6.4 Editing Label Files
context markers: 6.2.1 HTK Label Files
deleting: 6.4 Editing Label Files
editing: 6.4 Editing Label Files
external formats: 6.2 Label File Formats
merging: 6.4 Editing Label Files
moving level: 6.4 Editing Label Files
multiple level: 6.1 Label File Structure
replacing: 6.4 Editing Label Files
side-by-side: 6.1 Label File Structure
sorting: 6.4 Editing Label Files

language model scaling

12.7 N-Best Lists and Lattices

language models

bigram: 11.4 Bigram Language Models

lattice

comment lines: 16.2 Format
field names: 16.2 Format
format: 2.3.3 Recognition Tools, 16.2 Format
format: 2.3.3 Recognition Tools, 16.2 Format
header: 16.1 SLF Files
language model scale factor: 12.7 N-Best Lists and Lattices
link: 16.1 SLF Files
N-best: 1.6 Continuous Speech Recognition
node: 16.1 SLF Files
rescoring: 1.6 Continuous Speech Recognition
syntax: 16.3 Syntax

lattice generation

lattices

16.1 SLF Files

library modules

1.2 Isolated Word Recognition

likelihood computation

linear prediction

cepstra: 5.3 Linear Prediction Analysis

LINEIN

LINEOUT

3.5 Running the Recogniser Live

live input

<LLTCovar>

log arithmetic

LOPASS

LPC

LPCEPSTRA

LPCORDER

LPREFC

LS command

3.3.2 Step 10 - Making Tied-State Triphones, 3.5 Running the Recogniser Live, 9.5 Tree-Based Clustering

LT command

M-heaps

3.2.2 Step 7 - Fixing the Silence Models, 7.3 Macro Definitions

macro definition

7.3 Macro Definitions

macro substitution

7.3 Macro Definitions

macros

special meanings: 7.3 Macro Definitions
types: 7.3 Macro Definitions

marking word boundaries

9.2 Constructing Context-Dependent Models

master label files

3.1.4 Step 4 - Creating the Transcription Files, 6 Transcriptions and Label Files, 6.3.1 General Principles of MLFs, 8.2 Initialisation using HINIT, 8.5 Embedded Training using HEREST

embedded label definitions: 6.3.1 General Principles of MLFs
examples: 6.3.4 MLF Examples
multiple search paths: 6.3.1 General Principles of MLFs
pattern matching: 6.3.3 MLF Search
patterns: 3.1.4 Step 4 - Creating the Transcription Files, 6.3.3 MLF Search
patterns: 3.1.4 Step 4 - Creating the Transcription Files, 6.3.3 MLF Search
search: 6.3.3 MLF Search
sub-directory search: 6.3.3 MLF Search
syntax: 6.3.2 Syntax and Semantics
wildcards: 6.3.2 Syntax and Semantics

master macro file

3.2.1 Step 6 - Creating Flat Start Monophones

master macro files

input/output: 9.1 Using HHED

matrix dimensions

10.2 Using Discrete Models with Speech

MAXCLUSTITER

maximum model limit

MAXTRYOPEN

ME command

<Mean>

mean vector

MEASURESIL

mel scale

MELSPEC

memory

allocators: 4.7 Memory Management
element sizes: 4.7 Memory Management
statistics: 4.7 Memory Management

memory management

3.1.5 Step 5 - Coding the Data, 7.2 Basic HMM Definitions

MFCC coefficients

MICIN

minimum occupancy

9.6 Mixture Incrementing, 10.3 Tied Mixture Systems, 10.4 Parameter Smoothing

MINMIX

<Mixture>

7.2 Basic HMM Definitions, 7.9 The HMM Definition Language

mixture component

mixture incrementing

mixture splitting

10.3 Tied Mixture Systems

mixture tying

mixture weight floor

ML command

3.1.4 Step 4 - Creating the Transcription Files, 6 Transcriptions and Label Files

MLF

MMF

3.2.1 Step 6 - Creating Flat Start Monophones, 7.4 HMM Sets

model compaction

8.2 Initialisation using HINIT, 8.5 Embedded Training using HEREST

model training

clustering: 9.4 Data-Driven Clustering
compacting: 9.4 Data-Driven Clustering
context dependency: 9.2 Constructing Context-Dependent Models
embedded: 8.5 Embedded Training using HEREST
embedded subword formulae: 8.7.4 Embedded Model Reestimation(HEREST)
forward/backward formulae: 8.7.2 Forward/Backward Probabilities
HMM editing: 9.1 Using HHED
in parallel: 8.5 Embedded Training using HEREST
initialisation: 8.2 Initialisation using HINIT
isolated unit formulae: 8.7.3 Single Model Reestimation(HREST)
mixture components: 8.2 Initialisation using HINIT
pruning: 8.5 Embedded Training using HEREST
re-estimation formulae: 8.7 Parameter Re-Estimation Formulae
sub-word initialisation: 8.2 Initialisation using HINIT
tying: 9.3 Parameter Tying and Item Lists
update control: 8.2 Initialisation using HINIT
Viterbi formulae: 8.7.1 Viterbi Training (HINIT)
whole word: 8.2 Initialisation using HINIT

monitoring convergence

monophone HMM

construction of: 3.2 Creating Monophone HMMs

MP command

MT command

MU command

12.7 N-Best Lists and Lattices

mu law encoded files

5.8.4 NIST File Format

multiple alternative transcriptions

multiple hypotheses

16.1 SLF Files

multiple recognisers

12.2 Decoder Organisation

multiple streams

5.10 Multiple Input Streams

rules for: 5.10 Multiple Input Streams

multiple-tokens

1.6 Continuous Speech Recognition, 12.7 N-Best Lists and Lattices

N-best

N-grams

11.4 Bigram Language Models

NATURALREADORDER

4.9 Byte-swapping of HTK data files

NATURALWRITEORDER

4.9 Byte-swapping of HTK data files

NC command

network type

11 NetworksDictionaries and Language Models

networks

in recognition: 11.1 How Networks are Used
word-internal: 3.4.1 Step 11 - Recognising the Test Data

new features

in Version 2.0: 2.4 Whats New in Version 2.0?
in Version 2.1: 2.4.1 Whats New in Version 2.1?

NIST

2.3.4 Analysis Tool

NIST format

NIST scoring software

NIST Sphere data format

5.8.4 NIST File Format

non-emitting states

5.8.11 ALIEN and NOHEAD File Formats

non-printing chars

4.6 Strings and Names

NSAMPLES

NUMCEPS

5.3 Linear Prediction Analysis, 5.4 Filterbank Analysis

NUMCHANS

4.3 Configuration Files, 5.4 Filterbank Analysis

<NumMixes>

7.2 Basic HMM Definitions, 7.6 Discrete Probability HMMs

<NumStates>

7.9 The HMM Definition Language

observations

displaying structure of: 5.12 Viewing Speech with HLIST

operating system

4 The Operating Environment

outlier threshold

output filter

12.7 N-Best Lists and Lattices

output lattice format

output probability

continuous case: 1.3 Output Probability Specification, 7.1 The HMM Parameters
continuous case: 1.3 Output Probability Specification, 7.1 The HMM Parameters
discrete case: 7.1 The HMM Parameters

OUTSILWARN

8.4 Isolated Unit Re-Estimation using HREST

over-short training segments

parameter estimation

8 HMM Parameter Estimation

parameter kind

5.7.1 HTK Format Parameter Files

parameter tie points

7.3 Macro Definitions

parameter tying

3.3.1 Step 9 - Making Triphones from Monophones

parameterisation

3.1.5 Step 5 - Coding the Data

partial results

1.5 Recognition and Viterbi Decoding, 1.5 Recognition and Viterbi Decoding

path

as a token: 1.6 Continuous Speech Recognition

phone alignment

3.2.3 Step 8 - Realigning the Training Data

phone mapping

3.2.3 Step 8 - Realigning the Training Data

phone model initialisation

11.9 Other Kinds of Recognition System

phone recognition

phones

PHONESOUT

phonetic questions

4 The Operating Environment, 4.8 Input/Output via Pipes and Networks

pipes

PLAINHS

pre-emphasis

PREEMCOEF

2.3.2 Training Tools, 3.2.1 Step 6 - Creating Flat Start Monophones, 12.1 Decoder Operation, 12.3 Recognition using Test Databases

prompt script

generationof: 3.1.3 Step 3 - Recording the Data

prototype definition

2.3.2 Training Tools

pruning

in tied mixtures: 10.3 Tied Mixture Systems

pruning errors

8.5 Embedded Training using HEREST

QS command

3.3.2 Step 10 - Making Tied-State Triphones, 9.5 Tree-Based Clustering

qualifiers

5.1 General Mechanism, 5.7.1 HTK Format Parameter Files

_A: 5.6 Delta and Acceleration Coefficients
_C: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_C: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_C: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_D: 5.6 Delta and Acceleration Coefficients
_E: 5.5 Energy Measures
_K: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_K: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_K: 5.7.1 HTK Format Parameter Files, 5.7.1 HTK Format Parameter Files, 5.15 Summary
_N: 5.6 Delta and Acceleration Coefficients, 5.10 Multiple Input Streams
_N: 5.6 Delta and Acceleration Coefficients, 5.10 Multiple Input Streams
_O: 5.5 Energy Measures
_V: 5.11 Vector Quantisation, 5.13 Copying and Coding using HCOPY, 10.2 Using Discrete Models with Speech
_V: 5.11 Vector Quantisation, 5.13 Copying and Coding using HCOPY, 10.2 Using Discrete Models with Speech
_V: 5.11 Vector Quantisation, 5.13 Copying and Coding using HCOPY, 10.2 Using Discrete Models with Speech
_Z: 5.4 Filterbank Analysis
codes: 5.7.1 HTK Format Parameter Files
ESIG field specifiers: 5.7.2 Esignal Format Parameter Files
summary: 5.15 Summary

RAWENERGY

RE command

3.2.3 Step 8 - Realigning the Training Data

realignment

recogniser evaluation

3.4 Recogniser Evaluation

recogniser performance

3.1.3 Step 3 - Recording the Data

recognition

direct audio input: 3.5 Running the Recogniser Live
errors: 12.4 Evaluating Recognition Results
hypothesis: 12.1 Decoder Operation
network: 12.1 Decoder Operation
output: 3.5 Running the Recogniser Live
overall process: 11.1 How Networks are Used
results analysis: 3.4.1 Step 11 - Recognising the Test Data
statistics: 12.4 Evaluating Recognition Results
tools: 2.3.3 Recognition Tools

recording speech

RECOUTPREFIX

RECOUTSUFFIX

reflection coefficients

regression formula

removing outliers

3.3.2 Step 10 - Making Tied-State Triphones, 9.1 Using HHED, 9.4 Data-Driven Clustering

results analysis

2.3.4 Analysis Tool

RO command

RP command

RT command

5.11 Vector Quantisation, 10.2 Using Discrete Models with Speech

SAVEASVQ

SAVEBINARY

7.8 Binary Storage Format, 9.2 Constructing Context-Dependent Models

SAVECOMPRESSED

SAVEWITHCRC

4.2 Script Files, 8.2 Initialisation using HINIT

script files

search errors

12.1 Decoder Operation

segmental k-means

2.3.2 Training Tools

sentence generation

11.6 Testing a Word Network using HSGEN

SH command

SHAREDHS

3.2.2 Step 7 - Fixing the Silence Models

short pause

signals

for recording control: 12.6 Recognition using Direct Audio Input

silence floor

3.2.2 Step 7 - Fixing the Silence Models, 3.2.3 Step 8 - Realigning the Training Data, 9.3 Parameter Tying and Item Lists

silence model

SILFLOOR

simple differences

SIMPLEDIFFS

8.6 Single-Pass Retraining

single-pass retraining

singleton clusters

SK command

2.3.3 Recognition Tools, 3.1.1 Step 1 - the Task Grammar, 11 NetworksDictionaries and Language Models, 11.2 Word Networks and Standard Lattice Format

SLF

arc probabilities: 11.2 Word Networks and Standard Lattice Format
format: 11.2 Word Networks and Standard Lattice Format
null nodes: 11.2 Word Networks and Standard Lattice Format
word network: 11.2 Word Networks and Standard Lattice Format

SO command

software architecture

5.7.2 Esignal Format Parameter Files, 5.8 Waveform File Formats

SOURCEFORMAT

SOURCEKIND

5.1 General Mechanism, 12.6 Recognition using Direct Audio Input

SOURCELABEL

6.2 Label File Formats, 6.4 Editing Label Files

SOURCERATE

5.2 Speech Signal Processing, 5.9 Direct Audio Input/Output

SP command

speaker identifier

SPEAKEROUT

speech input

5 Speech Input/Output

automatic conversion: 5.1 General Mechanism
bandpass filtering: 5.4 Filterbank Analysis
blocking: 5.2 Speech Signal Processing
DC offset: 5.2 Speech Signal Processing
direct audio: 5.9 Direct Audio Input/Output
dynamic coefficents: 5.6 Delta and Acceleration Coefficients
energy measures: 5.5 Energy Measures
filter bank: 5.4 Filterbank Analysis
general mechanism: 5.1 General Mechanism
Hamming window function: 5.2 Speech Signal Processing
monitoring: 5.12 Viewing Speech with HLIST
pre-emphasis: 5.2 Speech Signal Processing
pre-processing: 5.2 Speech Signal Processing
summary of variables: 5.15 Summary
target kind: 5.1 General Mechanism

speech/silence detector

9.7 Miscellaneous Operations, 10.3 Tied Mixture Systems

speech;tex2html_html_special_mark_quot;input

byte order: 5.2 Speech Signal Processing

SS command

ST command

3.3.2 Step 10 - Making Tied-State Triphones, 9.5 Tree-Based Clustering

stacks

2.3.3 Recognition Tools, 3.1.1 Step 1 - the Task Grammar, 11 NetworksDictionaries and Language Models, 11.2 Word Networks and Standard Lattice Format

standard lattice format

definition;tex2html_html_special_mark_quot;: 16 HTK Standard Lattice Format (SLF)

standard options

4.4 Standard Options

-A: 4.4 Standard Options
-C: 4.3 Configuration Files, 4.4 Standard Options
-C: 4.3 Configuration Files, 4.4 Standard Options
-D: 4.4 Standard Options
-F: 5.7.2 Esignal Format Parameter Files, 5.8 Waveform File Formats, 6.2 Label File Formats
-F: 5.7.2 Esignal Format Parameter Files, 5.8 Waveform File Formats, 6.2 Label File Formats
-F: 5.7.2 Esignal Format Parameter Files, 5.8 Waveform File Formats, 6.2 Label File Formats
-G: 6.2 Label File Formats
-I: 6.3.1 General Principles of MLFs
-L: 6.3.1 General Principles of MLFs, 6.4 Editing Label Files, 8.2 Initialisation using HINIT
-L: 6.3.1 General Principles of MLFs, 6.4 Editing Label Files, 8.2 Initialisation using HINIT
-L: 6.3.1 General Principles of MLFs, 6.4 Editing Label Files, 8.2 Initialisation using HINIT
-S: 4.2 Script Files, 4.4 Standard Options, 6.3.1 General Principles of MLFs
-S: 4.2 Script Files, 4.4 Standard Options, 6.3.1 General Principles of MLFs
-S: 4.2 Script Files, 4.4 Standard Options, 6.3.1 General Principles of MLFs
-T: 4.4 Standard Options, 8.2 Initialisation using HINIT
-T: 4.4 Standard Options, 8.2 Initialisation using HINIT
-V: 4.4 Standard Options
summary: 4.10 Summary

state clustering

3.3.2 Step 10 - Making Tied-State Triphones, 9.4 Data-Driven Clustering

state transitions

adding/removing: 9.7 Miscellaneous Operations

state tying

statistics

state occupation: 3.3.1 Step 9 - Making Triphones from Monophones

statistics file

3.3.2 Step 10 - Making Tied-State Triphones, 9.1 Using HHED

statistics;tex2html_html_special_mark_quot;file

7.2 Basic HMM Definitions, 7.5 Tied-Mixture Systems

<Stream>

stream weight

1.3 Output Probability Specification, 7.1 The HMM Parameters

<StreamInfo>

7.2 Basic HMM Definitions, 7.9 The HMM Definition Language

streams

3.1.2 Step 2 - the Dictionary

stress marking

string matching

string values

strings

metacharacters in: 4.6 Strings and Names
output of: 4.6 Strings and Names
rules for: 4.6 Strings and Names

SU command

11.5 Building a Word Network with HBUILD, 16.1 SLF Files

sub-lattices

SUBLAT

11.5 Building a Word Network with HBUILD, 16.1 SLF Files

SW command

<SWeights>

4.3 Configuration Files, 5.1 General Mechanism, 10.2 Using Discrete Models with Speech

TARGETKIND

TARGETRATE

3.1.1 Step 1 - the Task Grammar, 11.3 Building a Word Network with HPARSE

task grammar

TB command

3.3.2 Step 10 - Making Tied-State Triphones, 9.5 Tree-Based Clustering

TC command

6.4 Editing Label Files, 9.2 Constructing Context-Dependent Models, 9.4 Data-Driven Clustering

tee-models

3.2.2 Step 7 - Fixing the Silence Models, 7.7 Tee Models

in networks: 11.3 Building a Word Network with HPARSE

termination

3.2.2 Step 7 - Fixing the Silence Models, 9.3 Parameter Tying and Item Lists

TI command

tied parameters

9.3 Parameter Tying and Item Lists

tied-mixture system

7.5 Tied-Mixture Systems

tied-mixtures

9.3 Parameter Tying and Item Lists, 10.3 Tied Mixture Systems

build procedure: 10.3 Tied Mixture Systems
output distribution: 7.5 Tied-Mixture Systems

tied-state

3.1.2 Step 2 - the Dictionary, 6.4 Editing Label Files

TIMIT database

token history

12.1 Decoder Operation

token passing

12.1 Decoder Operation

Token Passing Model

1.4 Baum-Welch Re-Estimation, 1.5 Recognition and Viterbi Decoding

total likelihood

TR command

6.1 Label File Structure, 6.4 Editing Label Files

tracing

4.4 Standard Options

training

sub-word: 8.1 Training Strategies
whole-word: 8.1 Training Strategies

training tools

2.3.2 Training Tools

TRANSALT

transcription

orthographic: 3.1.4 Step 4 - Creating the Transcription Files

transcriptions

model level: 12.5 Generating Forced Alignments
phone level: 12.5 Generating Forced Alignments
word level: 12.5 Generating Forced Alignments

transitions

adding them: 3.2.2 Step 7 - Fixing the Silence Models

TRANSLEV

6.1 Label File Structure, 6.4 Editing Label Files

<TransP>

tree building

tree optimisation

triphones

by cloning: 3.3.1 Step 9 - Making Triphones from Monophones
from monophones: 3.3.1 Step 9 - Making Triphones from Monophones
notation: 3.3.1 Step 9 - Making Triphones from Monophones
synthesising unseen: 3.5 Running the Recogniser Live
word internal: 3.3.1 Step 9 - Making Triphones from Monophones

tying

examples of: 9.3 Parameter Tying and Item Lists
exemplar selection: 9.3 Parameter Tying and Item Lists
states: 3.3.2 Step 10 - Making Tied-State Triphones
transition;tex2html_html_special_mark_quot;matrices: 3.3.1 Step 9 - Making Triphones from Monophones

UF command

9.1 Using HHED

under-training

10.4 Parameter Smoothing

uniform segmentation

3.3.2 Step 10 - Making Tied-State Triphones, 9.5 Tree-Based Clustering

unseen triphones

synthesising: 9.5 Tree-Based Clustering

up-mixing

upper triangular form

<Use> clause (V1.5)

USEHAMMING

USEPOWER

5.9 Direct Audio Input/Output, 12.6 Recognition using Direct Audio Input

USESILDET

UT command

9.3 Parameter Tying and Item Lists

V1.5 grammar files

2.4 Whats New in Version 2.0?, 5.6 Delta and Acceleration Coefficients, 11.3 Building a Word Network with HPARSE

V1COMPAT

varFloorN

<Variance>

3.2.1 Step 6 - Creating Flat Start Monophones

flooring problems: 3.3.2 Step 10 - Making Tied-State Triphones

variance floor macros

generating: 8.3 Flat Starting with HCOMPV

variance floors

7.2 Basic HMM Definitions, 7.9 The HMM Definition Language

<VecSize>

vector dimensions

vector quantisation

code book external format: 5.11 Vector Quantisation
distance metrics: 5.11 Vector Quantisation
type of: 5.11 Vector Quantisation
uses;tex2html_html_special_mark_quot;of: 5.11 Vector Quantisation

Version 1.5

1.4 Baum-Welch Re-Estimation, 8.2 Initialisation using HINIT

Viterbi training

VQ codebook

VQTABLE

warning codes

full listing: 15 Error and Warning Codes

warning message

format: 15 Error and Warning Codes

warnings

waveform capture

3.3.1 Step 9 - Making Triphones from Monophones, 9.2 Constructing Context-Dependent Models

WB command

whole word modelling

11.9 Other Kinds of Recognition System

whole word recognition

WINDOWSIZE

4.3 Configuration Files, 5.2 Speech Signal Processing

WLR

word equivalence

3.4.1 Step 11 - Recognising the Test Data

word insertion penalty

word internal

3.3.1 Step 9 - Making Triphones from Monophones

Word Link Record

3.1.2 Step 2 - the Dictionary

word list

word N-best

expansion rules

11.9 Other Kinds of Recognition System

word networks

tee-models in: 11.3 Building a Word Network with HPARSE

word spotting

word-end nodes

1.6 Continuous Speech Recognition, 1.6 Continuous Speech Recognition, 12.1 Decoder Operation

word-internal network expansion

11.5 Building a Word Network with HBUILD

word-loop network

word-pair grammar

11.5 Building a Word Network with HBUILD

ZMEANSOURCE