/// <summary>
/// Dump the data in the phone.
/// </summary>
/// <param name="scriptSyllable">The script syllable to store the data dumped from the phones.</param>
/// <param name="utt">The utterance.</param>
/// <param name="syllable">The syllable which contains these phones.</param>
/// <param name="phoneIndex">Phone index to mark the phone in the Utt.Phones.</param>
/// <param name="unitIndex">Unit index to mark the unit in the Utt.Units.</param>
/// <param name="f0StartIndex">F0 index to mark the start position in the F0s.</param>
/// <param name="ttsEngine">The object ttsEngine to help to convert the Pos and get sentence id.</param>
private static void DumpPhones(ScriptSyllable scriptSyllable, SP.TtsUtterance utt,
SP.TtsSyllable syllable, ref int phoneIndex, ref int unitIndex, ref int f0StartIndex, SP.TtsEngine ttsEngine)
{
Debug.Assert(scriptSyllable != null, "ScriptSyllable should not be null");
Debug.Assert(utt != null, "Utt should not be null");
Debug.Assert(syllable != null, "Syllable should not be null");
Debug.Assert(phoneIndex >= 0, "PhoneIndex should not be less than 0");
Debug.Assert(f0StartIndex >= 0, "f0StartIndex should not be less than 0");
Debug.Assert(ttsEngine != null, "ttsEngine should not be null");
WuiManager wuiManager = null;
if (utt.Segments.Count > 0)
{
int bestNodeIndex = (int)utt.UnitLattice.WucList[unitIndex].BestNodeIndex;
wuiManager = ttsEngine.RUSVoiceDataManager.GetWuiManagerByUnitCostNode(utt.UnitLattice.WucList[unitIndex].WucNodeList[bestNodeIndex]);
}
// Go through each phone in the syllable.
SP.TtsPhone phone = syllable.FirstPhone;
while (phone != null)
{
// Dump the pronunciation of the phone.
string phonePronunciation = Pronunciation.RemoveStress(phone.Pronunciation.ToLowerInvariant()).Trim();
// Remove the tone from the phone pronunciation if it exist.
if (phone.Tone != 0)
{
phonePronunciation = Pronunciation.RemoveTone(phonePronunciation).Trim();
}
ScriptPhone scriptPhone = new ScriptPhone(phonePronunciation);
scriptPhone.Tone = phone.Tone.ToString();
scriptPhone.Stress = (TtsStress)phone.Stress;
if (phone.Pronunciation != PronOfSilence)
{
if (wuiManager != null)
{
scriptPhone.SentenceId = wuiManager.GetSentenceId(utt.Segments[unitIndex].WaveUnitInfo);
}
if (phone.Unit != null)
{
scriptPhone.UnitIndex = (int)phone.Unit.UnitIndex;
}
}
scriptPhone.Acoustics = new ScriptAcoustics();
// Dump the segments.
if (utt.Segments.Count > 0 && !utt.Segments[unitIndex].Unit.UnitText.Equals(PronOfSilence)
&& !utt.Segments[unitIndex].Unit.UnitText.Equals(PronOfShortPause))
{
scriptPhone.Acoustics.Duration = (int)utt.Segments[unitIndex].WaveUnitInfo.WaveLength + (int)utt.Segments[unitIndex + 1].WaveUnitInfo.WaveLength;
int segStart = (int)utt.Segments[unitIndex].WaveUnitInfo.RecordingWaveStartPosition;
int segEnd = segStart + (int)utt.Segments[unitIndex].WaveUnitInfo.WaveLength;
scriptPhone.Acoustics.SegmentIntervals.Add(new SegmentInterval(segStart, segEnd));
segStart = (int)utt.Segments[unitIndex + 1].WaveUnitInfo.RecordingWaveStartPosition;
segEnd = segStart + (int)utt.Segments[unitIndex + 1].WaveUnitInfo.WaveLength;
scriptPhone.Acoustics.SegmentIntervals.Add(new SegmentInterval(segStart, segEnd));
}
// Relative begin position of the uvsegment interval.
int relativeBegin = 0;
// Relative end position of the uvsegment interval.
int relativeEnd = 0;
// When go through the F0 values, this valuie to identify if meet the first voiced segment.
bool reBeginPositionFindOut = false;
// Check if all the F0 values in one state are equals to 0. If yes, don't write down the uvseg.
bool isF0ValueExist = false;
// Dump the durations and F0s in each state.
if (utt.Acoustic.Durations != null)
{
for (int i = 0; i < utt.Acoustic.Durations[phoneIndex].Length; ++i)
{
ScriptState scriptState = new ScriptState();
// Dump duration
int durationInFrame = (int)utt.Acoustic.Durations[phoneIndex][i];
scriptState.Acoustics = new ScriptAcoustics(durationInFrame * MillisecondsPerFrame);
// Dump F0s
if (utt.Acoustic.F0s != null)
{
ScriptUvSeg scriptUvSeg = GetF0Contour(utt, f0StartIndex, durationInFrame, ScriptAcousticChunkEncoding.Text,
ref relativeBegin, ref relativeEnd, ref reBeginPositionFindOut, ref isF0ValueExist);
if (isF0ValueExist == true)
{
scriptState.Acoustics.UvSegs.Add(scriptUvSeg);
}
f0StartIndex += durationInFrame;
}
scriptPhone.States.Add(scriptState);
}
}
// Dump the uvsegment relative interval.
if (utt.Acoustic.F0s != null && !phone.Pronunciation.Equals(PronOfSilence)
&& !phone.Pronunciation.Equals(PronOfShortPause))
{
ScriptUvSeg uvSegForRelativeInterval = new ScriptUvSeg(ScriptUvSegType.Mixed);
uvSegForRelativeInterval.Interval = new ScriptUvSegInterval(relativeBegin * 5, relativeEnd * 5);
scriptPhone.Acoustics.UvSegs.Add(uvSegForRelativeInterval);
}
phoneIndex++;
unitIndex++;
if (wuiManager != null &&
!phone.Pronunciation.Equals(PronOfSilence) &&
!phone.Pronunciation.Equals(PronOfShortPause))
{
// if it is not an silence phone, the according unit must be an half phone unit,
// we need skip the right half phone to move next phone's unit
unitIndex++;
}
scriptSyllable.Phones.Add(scriptPhone);
if (phone == syllable.LastPhone)
{
break;
}
phone = phone.Next;
}
}
/// <summary>
/// Load state from XmlTextReader.
/// </summary>
/// <param name="reader">Xml text reader.</param>
/// <returns>Script state.</returns>
private static ScriptState LoadState(XmlTextReader reader)
{
Debug.Assert(reader != null);
ScriptState state = new ScriptState();
if (!reader.IsEmptyElement)
{
while (reader.Read())
{
if (reader.NodeType == XmlNodeType.Element && reader.Name == "acoustics")
{
state.Acoustics = new ScriptAcoustics();
state.Acoustics.ParseFromXml(reader);
}
else if (reader.NodeType == XmlNodeType.EndElement && reader.Name == "state")
{
break;
}
}
}
return state;
}
