public LinguistciFeatureExtractor(IEnumerable<LinguisticFeatureInfo> featureInfos, TtsPhoneSet phoneSet,
TtsPosSet posSet, CustomizedFeaturePluginManager manager, ILogger logger)
{
if (featureInfos == null)
{
throw new ArgumentNullException("featureInfos");
}
if (phoneSet == null)
{
throw new ArgumentNullException("phoneSet");
}
if (posSet == null)
{
throw new ArgumentNullException("posSet");
}
Logger = logger ?? new NullLogger();
PhoneSet = phoneSet;
PosSet = posSet;
// Builds the name of feature name set.
FeatureNameSetName = Helper.NeutralFormat("VoiceModelTrainer.{0}", PhoneSet.Language.ToString());
try
{
// Creates a feature extration engine.
ExtractionEngine = new FeatureExtractionEngine();
// Creates the feature meta data.
FeatureMetas = ExtractionEngine.Convert(LabelFeatureNameSet.MandatoryFeatureNames.ToList());
FeatureInfos = new List<LinguisticFeatureInfo>();
FeatureValueRecords = new List<FeatureValueRecord>();
for (int i = 0; i < LabelFeatureNameSet.MandatoryFeatureNames.Length; ++i)
{
FeatureInfos.Add(null);
FeatureValueRecords.Add(new FeatureValueRecord());
}
foreach (LinguisticFeatureInfo info in featureInfos)
{
FeatureValueRecords.Add(new FeatureValueRecord());
int index = Array.IndexOf(LabelFeatureNameSet.MandatoryFeatureNames, info.Name);
if (index < 0)
{
FeatureMetas.Add(ExtractionEngine.Convert(info.Name, info.ExtendedProperty));
FeatureInfos.Add(info);
}
else
{
FeatureInfos[index] = info;
}
}
}
catch (EspException e)
{
throw new InvalidDataException("Feature extraction engine error", e);
}
// Checks whether need pos and ToBI accent.
for (int i = 0; i < FeatureMetas.Count; ++i)
{
if (!NeedPos &&
(FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_POS ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_POSTAGGER_POS))
{
NeedPos = true;
}
if (!NeedToBI &&
(FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_PRIMARY_ACCENT_POSITION ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_SYLL_NUM_FROM_LEFT_ACCENT ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_SYLL_NUM_TO_RIGHT_ACCENT ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_ACCENTED_SYLL_NUM_BEFORE_CURR_SYLL ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_ACCENTED_SYLL_NUM_AFTER_CURR_SYLL ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_ACCENT ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_TOBI_ACCENT ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_IS_ACCENTED ||
FeatureMetas[i].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_TOBI_FINAL_BOUNDARY_TONE))
{
NeedToBI = true;
}
}
// Gets Phoneme according to phone set.
MemoryStream phoneStream = new MemoryStream();
ErrorSet errorSet = PhoneSetCompiler.Compile(phoneSet, phoneStream);
if (errorSet.Count > 0)
{
foreach (Error error in errorSet.Errors)
{
Logger.LogLine(error.ToString());
}
if (errorSet.Contains(ErrorSeverity.MustFix))
{
throw new InvalidDataException("Error happens in tts phone set compiling");
}
}
phoneStream.Seek(0, SeekOrigin.Begin);
Phoneme = new Phoneme(phoneStream, (Language)phoneSet.Language);
// Gets the utterance extenders.
if (manager != null)
{
List<PluginInfo> pluginInfos = manager.GetPlugins(CustomizedFeaturePluginManager.AttachBeforeExtraction);
if (pluginInfos != null)
{
UtteranceExtenders = UtteranceExtenderFinder.LoadUtteranceExtenders(pluginInfos);
}
}
// Initialize ZhToneIndexPlugin if the language is zh-CN
if (Language.ZhCN == (Language)phoneSet.Language)
{
ChineseToneIndexExtractor = new ChineseToneIndexExtractor();
ChineseToneIndexExtractor.Initialize(Phoneme);
}
// Creates feature name set.
if (LabelFeatureNameSet.Exist(FeatureNameSetName))
{
FeatureNameSet = LabelFeatureNameSet.Query(FeatureNameSetName);
}
else
{
FeatureNameSet = LabelFeatureNameSet.Create(FeatureNameSetName,
FeatureMetas.Select(o => o.Name).ToList());
}
}
/// <summary>
/// Initializes a new instance of the <see cref="UtteranceBuilder"/> class.
/// </summary>
/// <param name="phoneSet">
/// The phone set.
/// </param>
/// <param name="posSet">
/// The pos set.
/// </param>
/// <param name="phoneme">
/// The phoneme.
/// </param>
/// <exception cref="ArgumentNullException">
/// Exception.
/// </exception>
public UtteranceBuilder(TtsPhoneSet phoneSet, TtsPosSet posSet, Phoneme phoneme)
{
if (phoneSet == null)
{
throw new ArgumentNullException("phoneSet");
}
if (posSet == null)
{
throw new ArgumentNullException("posSet");
}
if (phoneme == null)
{
throw new ArgumentNullException("phoneme");
}
PhoneSet = phoneSet;
PosSet = posSet;
Phoneme = phoneme;
NeedPunctuation = false;
}
