/// <summary>
/// Extracts the features of the given utterance.
/// </summary>
/// <param name="sentId">
/// Sentence id.
/// </param>
/// <param name="utterance">
/// Service Provider utterance object.
/// </param>
/// <returns>
/// The sentence contains all the features.
/// </returns>
/// <exception cref="InvalidDataException">
/// Exception.
/// </exception>
public Sentence Extract(string sentId, TtsUtterance utterance)
{
List<FeatureVector> vectors;
try
{
// Then, extracts the features.
vectors = ExtractionEngine.Extract(utterance, FeatureMetas);
}
catch (EspException e)
{
throw new InvalidDataException(Helper.NeutralFormat("Extract feature error on sentence \"{0}\"",
sentId), e);
}
// Validates the extracted vectors.
if (vectors.Count != FeatureMetas.Count)
{
throw new InvalidDataException(
Helper.NeutralFormat("Length of result is mismatch on sentence \"{0}\"", sentId));
}
for (int i = 0; i < vectors.Count; i++)
{
if (vectors[i].Count != utterance.Phones.Count)
{
throw new InvalidDataException(
Helper.NeutralFormat("Length of vector is mismatch on sentence \"{0}\"", sentId));
}
}
// Creates a sentence to store all the features.
Sentence sentence = new Sentence { Id = sentId };
for (int i = 0; i < vectors[0].Count; ++i)
{
// Create candidates for each phoneme.
PhoneSegment p = new PhoneSegment
{
Sentence = sentence,
Index = i,
Features = vectors.Select(v => v[i])
.Skip(LabelFeatureNameSet.MandatoryFeatureNames.Length).ToArray(),
};
// Create the label to store the features.
Label label = new Label(FeatureNameSet);
for (int j = 0; j < vectors.Count; ++j)
{
if (vectors[j][i].ValueType == FeatureValueType.FEATURE_VALUE_TYPE_UNKOWN)
{
label.SetFeatureValue(FeatureNameSet.FeatureNames[j], Label.NotApplicableFeatureValue);
}
else if (FeatureMetas[j].Property == TtsFeatureProperty.TTS_FEATURE_PROPERTY_PHONE_ID)
{
Phone phone = PhoneSet.GetPhone(vectors[j][i].IntValue);
label.SetFeatureValue(FeatureNameSet.FeatureNames[j], Offline.Phoneme.ToHtk(phone.Name));
}
else
{
label.SetFeatureValue(FeatureNameSet.FeatureNames[j],
vectors[j][i].IntValue.ToString(CultureInfo.InvariantCulture));
}
// Updates the corresponding value records.
FeatureValueRecords[j].Update(vectors[j][i]);
}
p.Label = label;
sentence.PhoneSegments.Add(p);
}
return sentence;
}
/// <summary>
/// Initializes a new instance of the <see cref="PhoneCandidate"/> class.
/// </summary>
/// <param name="phoneme">Phone segment link.</param>
public PhoneCandidate(PhoneSegment phoneme)
{
Type = UnitCandidateType.Phone;
Id = InvalidId;
Label = phoneme.Label;
Name = phoneme.Name;
LeftPhoneme = phoneme.LeftPhoneme;
RightPhoneme = phoneme.RightPhoneme;
StartTime = phoneme.StartTime;
EndTime = phoneme.EndTime;
Index = phoneme.Index;
IndexOfNonSilence = phoneme.IndexOfNonSilence;
MustHold = false;
PhoneSegmentLink = phoneme;
Sentence = phoneme.Sentence;
SilenceCandidate = phoneme.Name == Phoneme.ToHtk(Phoneme.SilencePhone);
}
/// <summary>
/// Initializes a new instance of the <see cref="HalfPhoneCandidate"/> class.
/// </summary>
/// <param name="phoneme">Phone segment link.</param>
/// <param name="isLeftHalfPhone">Flag to left or right half phone.</param>
public HalfPhoneCandidate(PhoneSegment phoneme, bool isLeftHalfPhone)
{
Type = UnitCandidateType.Halfphone;
Id = InvalidId;
Label = phoneme.Label;
// bug #93090, TBD: Offline should call serviceProvider to get half phone name and boundary
IsLeftHalfPhone = isLeftHalfPhone;
Name = isLeftHalfPhone ? "hpl_" + phoneme.Name : "hpr_" + phoneme.Name;
int middleTime = (int)(((phoneme.StateAlignments[2].StartTime + phoneme.StateAlignments[2].EndTime) / 2 / 50000.0) + 0.6) * 50000;
StartTime = isLeftHalfPhone ? phoneme.StartTime : middleTime;
EndTime = isLeftHalfPhone ? middleTime : phoneme.EndTime;
LeftPhoneme = phoneme.LeftPhoneme;
RightPhoneme = phoneme.RightPhoneme;
Index = isLeftHalfPhone ? phoneme.Index * 2 : (phoneme.Index * 2) + 1;
IndexOfNonSilence = isLeftHalfPhone ? phoneme.IndexOfNonSilence * 2 : (phoneme.IndexOfNonSilence * 2) + 1;
MustHold = false;
PhoneSegmentLink = phoneme;
Sentence = phoneme.Sentence;
SilenceCandidate = phoneme.Name == Phoneme.ToHtk(Phoneme.SilencePhone);
}
/// <summary>
/// Loads one sentence from the master label file.
/// Please notice this load may be called multi-times to load the different
/// Information, such as full-context label, alignment data and so on.
/// </summary>
/// <param name="reader">StreamReader of master label file.</param>
/// <returns>A bool value indicates whether end of sentence exists.</returns>
public bool Load(StreamReader reader)
{
// Load all the lines belong to this sentence.
bool endOfSentenceExist = false;
List<string> lines = LoadLines(reader, ref endOfSentenceExist);
int indexOfLines = 0;
if (_phoneSegments.Count == 0)
{
// The first time to load the phone segment.
int index = 0;
int indexOfNonSilence = 0;
while (indexOfLines < lines.Count)
{
PhoneSegment phoneSegment = new PhoneSegment();
phoneSegment.Load(lines, ref indexOfLines);
phoneSegment.Index = index++;
phoneSegment.IndexOfNonSilence = (!Phoneme.IsSilenceFeature(phoneSegment.Name)) ? indexOfNonSilence++ : -1;
phoneSegment.Sentence = this;
_phoneSegments.Add(phoneSegment);
}
}
else
{
// Reload information about the phone segment.
try
{
foreach (PhoneSegment phoneSegment in _phoneSegments)
{
phoneSegment.Load(lines, ref indexOfLines);
}
}
catch (InvalidDataException e)
{
throw new InvalidDataException("Mismatched data between multi master label files", e);
}
}
if (indexOfLines != lines.Count)
{
throw new InvalidDataException("Mismatched data between multi master label files");
}
return endOfSentenceExist;
}