marytts.modules

Class HTSEngine

java.lang.Object

marytts.modules.InternalModule

marytts.modules.HTSEngine

All Implemented Interfaces:

MaryModule

public class HTSEngine
extends InternalModule

HTSEngine: a compact HMM-based speech synthesis engine. Java port and extension of HTS engine API version 1.04 Extension: mixed excitation

Author:

Marc Schröder, Marcela Charfuelan

Field Summary

Fields inherited from class marytts.modules.InternalModule

logger, state

Fields inherited from interface marytts.modules.MaryModule

MODULE_OFFLINE, MODULE_RUNNING

Constructor Summary

Constructors

Constructor and Description
HTSEngine()

Method Summary

Methods

Modifier and Type	Method and Description
Vector<HTSEngineTest.PhonemeDuration>	getAlignDurations()
double	getNewStateDurationFactor()
boolean	getPhonemeAlignmentForDurations()
String	getRealisedDurations()
boolean	getStateAlignmentForDurations()
static List<Target>	getTargets(Scanner s, HMMData htsData)
static List<Target>	getTargetsFromFile(String LabFile, HMMData htsData) Reads the Label file, the file which contains the Mary context features, creates an scanner object and calls getTargets
List<Target>	getTargetsFromText(String LabText, HMMData htsData) Creates a scanner object with the Mary context features contained in Labtext and calls getTargets
static void	main(String[] args) Stand alone testing using a TARGETFEATURES file as input.
void	powerOnSelfTest() This module is actually tested as part of the HMMSynthesizer test, for which reason this method does nothing.
MaryData	process(MaryData d, List<Target> targetFeaturesList, List<Element> segmentsAndBoundaries, List<Element> tokensAndBoundaries) This functions process directly the target features list: targetFeaturesList when using external prosody, duration and f0 are read from acoustparams: segmentsAndBoundaries realised durations and f0 are set in: tokensAndBoundaries when calling this function HMMVoice must be initialised already, that is TreeSet and ModelSet must be loaded already.
protected HTSUttModel	processTargetList(List<Target> targetFeaturesList, List<Element> segmentsAndBoundaries, HMMData htsData) Process feature vectors in target list to generate a list of models for generation and realisation
HTSUttModel	processUttFromFile(String feaFile, HMMData htsData)
void	setAlignDurations(Vector<HTSEngineTest.PhonemeDuration> val)
void	setNewStateDurationFactor(double dval)
void	setPhonemeAlignmentForDurations(boolean bval)
void	setRealisedDurations(String str)
static void	setRealisedProsody(List<Element> tokensAndBoundaries, HTSUttModel um)
void	setStateAlignmentForDurations(boolean bval)

Methods inherited from class marytts.modules.InternalModule

getInputType, getLocale, getOutputType, getState, inputType, name, outputType, process, shutdown, startup

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

HTSEngine

public HTSEngine()

Method Detail

getRealisedDurations

public String getRealisedDurations()

getPhonemeAlignmentForDurations

public boolean getPhonemeAlignmentForDurations()

getStateAlignmentForDurations

public boolean getStateAlignmentForDurations()

getAlignDurations

public Vector<HTSEngineTest.PhonemeDuration> getAlignDurations()

getNewStateDurationFactor

public double getNewStateDurationFactor()

setRealisedDurations

public void setRealisedDurations(String str)

setStateAlignmentForDurations

public void setStateAlignmentForDurations(boolean bval)

setPhonemeAlignmentForDurations

public void setPhonemeAlignmentForDurations(boolean bval)

setAlignDurations

public void setAlignDurations(Vector<HTSEngineTest.PhonemeDuration> val)

setNewStateDurationFactor

public void setNewStateDurationFactor(double dval)

powerOnSelfTest

public void powerOnSelfTest()
                     throws Error

This module is actually tested as part of the HMMSynthesizer test, for which reason this method does nothing.

Specified by:

powerOnSelfTest in interface MaryModule

Overrides:

powerOnSelfTest in class InternalModule

Throws:

Error - Error

process

public MaryData process(MaryData d,
               List<Target> targetFeaturesList,
               List<Element> segmentsAndBoundaries,
               List<Element> tokensAndBoundaries)
                 throws Exception

This functions process directly the target features list: targetFeaturesList when using external prosody, duration and f0 are read from acoustparams: segmentsAndBoundaries realised durations and f0 are set in: tokensAndBoundaries when calling this function HMMVoice must be initialised already, that is TreeSet and ModelSet must be loaded already.

Parameters:

d - : to get the default voice and locale

targetFeaturesList - : the actual input data to HTS based synthesis

segmentsAndBoundaries - : to update segment timings that are influenced by HMM state selection

tokensAndBoundaries - :

Returns:

output

Throws:

Exception - Exception

setRealisedProsody

public static void setRealisedProsody(List<Element> tokensAndBoundaries,
                      HTSUttModel um)
                               throws SynthesisException

Throws:

SynthesisException

processUttFromFile

public HTSUttModel processUttFromFile(String feaFile,
                             HMMData htsData)
                               throws Exception

Throws:

Exception

getTargetsFromFile

public static List<Target> getTargetsFromFile(String LabFile,
                              HMMData htsData)
                                       throws Exception

Reads the Label file, the file which contains the Mary context features, creates an scanner object and calls getTargets

Parameters:

LabFile - LabFile

htsData - htsData

Returns:

targets

Throws:

Exception - Exception

getTargetsFromText

public List<Target> getTargetsFromText(String LabText,
                              HMMData htsData)
                                throws Exception

Creates a scanner object with the Mary context features contained in Labtext and calls getTargets

Parameters:

LabText - LabText

htsData - htsData

Returns:

targets

Throws:

Exception - Exception

getTargets

public static List<Target> getTargets(Scanner s,
                      HMMData htsData)

processTargetList

protected HTSUttModel processTargetList(List<Target> targetFeaturesList,
                            List<Element> segmentsAndBoundaries,
                            HMMData htsData)
                                 throws Exception

Process feature vectors in target list to generate a list of models for generation and realisation

Parameters:

targetFeaturesList - : each target must contain the corresponding feature vector

segmentsAndBoundaries - : if applying external prosody provide acoust params as a list of elements

htsData - : parameters and configuration of the voice

Returns:

um

Throws:

Exception - Exception

main

public static void main(String[] args)
                 throws IOException,
                        InterruptedException,
                        Exception

Stand alone testing using a TARGETFEATURES file as input.

Parameters:

args - args

Throws:

IOException - IOException

InterruptedException - InterruptedException

Exception - Exception