/// <summary>
/// Estimates the phoneme-stops in a specified word. Each phoneme gets
/// a weight that will be interpreted as its duration within the word.
/// </summary>
/// <param name="ar"><see cref="OrthographicResult"/></param>
static void phStops(OrthographicResult ar)
{
//logfile.Log("phStops()");
var stops = new List <decimal>();
decimal tally = 0;
foreach (var phon in ar.Phons)
{
switch (phon)
{
// curious where 0100010 got these - intuition perhaps.
// TODO: French etc.
case "aa":
case "ae":
case "ah":
case "ax":
case "ay":
case "b":
case "eh":
case "l":
case "r":
case "w":
stops.Add(tally += 5);
break;
case "ao":
case "aw":
case "er":
case "ey":
case "ow":
case "oy":
case "uh":
case "uw":
stops.Add(tally += 6);
break;
default:
stops.Add(tally += 3);
break;
}
}
if (stops.Count != 0)
{
decimal factor = (ar.Stop - ar.Start) / tally;
int i = 0;
for (; i != stops.Count - 1; ++i)
{
ar.phStops.Add(OrthographicResult.Truncate(stops[i] * factor) + ar.Start);
}
ar.phStops.Add(ar.Stop); // ensure the final phoneme-stop IS the word-stop.
}
}
void LoadTable(string pfe)
{
_ars_alt = new List <OrthographicResult>();
using (var fs = new FileStream(pfe, FileMode.Open, FileAccess.Read, FileShare.Read))
{
var sr = new StreamReader(fs, Encoding.ASCII);
OrthographicResult ar = null;
string l;
while ((l = sr.ReadLine()) != null && l != String.Empty)
{
// TODO: WARNING Do error checks ->
string[] a = l.Split(DELI);
if (Utility.isWordstart(a[0]))
{
if (ar != null)
{
_ars_alt.Add(ar);
}
ar = new OrthographicResult();
ar.Orthography = String.Empty;
ar.Confi = 0f;
ar.Level = String.Empty;
ar.Phons = new List <string>();
ar.Phons.Add(a[1]);
ar.Start = Decimal.Parse(a[2]);
ar.Stop = Decimal.Parse(a[3]);
ar.phStops.Add(Decimal.Parse(a[3]));
}
else
{
ar.Phons.Add(a[1]);
ar.phStops.Add(Decimal.Parse(a[3]));
}
}
if (ar != null)
{
_ars_alt.Add(ar);
}
sr.Close();
}
if (_ars_alt.Count != 0)
{
AlternateData();
}
}
/// <summary>
/// Inserts any required silences in the phrase and estimates the
/// phoneme-stops in each word.
/// </summary>
void Orthography()
{
#if DEBUG
logfile.Log();
logfile.Log("Orthography() _generato= " + _generato);
#endif
List <OrthographicResult> ars = null;
switch (_generato)
{
case Generator.Dictati: ars = _ars_def; break;
case Generator.Dialogi: ars = _ars_enh; break;
}
OrthographicResult ar;
decimal stop = 0;
for (int i = 0; i != ars.Count; ++i)
{
if ((ar = ars[i]).Start > stop) // TODO: use a tolerance Eg. 10..15 millisec
{
#if DEBUG
logfile.Log(". . insert silence");
#endif
OrthographicResult sil = CreateSilence();
sil.Start = stop;
sil.phStops.Add(sil.Stop = ar.Start);
ars.Insert(i, sil);
++i;
}
#if DEBUG
logfile.Log(". ar.Orthography= " + ar.Orthography);
string phons = String.Empty;
foreach (var phon in ar.Phons)
{
if (phons != String.Empty)
{
phons += " ";
}
phons += phon;
}
logfile.Log(". ar.Phons= " + phons);
#endif
phStops(ar);
stop = ar.Stop;
}
}
/// <summary>
/// Creates a silence.
/// </summary>
/// <returns>an <see cref="OrthographicResult"/> w/ blank Orthography</returns>
static OrthographicResult CreateSilence()
{
var sil = new OrthographicResult();
sil.Orthography = String.Empty;
sil.Phons = new List <string>();
sil.Phons.Add(StaticData.SIL);
sil.Confi = 1f;
sil.Level = String.Empty;
return(sil);
}
/// <summary>
/// Builds a list of OrthographicResults from the edited DataTable.
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
void click_Accept(object sender, EventArgs e)
{
#if DEBUG
logfile.Log();
logfile.Log("click_Accept()");
#endif
// TODO: only if changed
_f._ars_alt = new List <OrthographicResult>();
OrthographicResult result;
bool decr;
for (int r = 0; r != _dt.Rows.Count; ++r)
{
string pos = _dt.Rows[r][0] as String; // pos
#if DEBUG
logfile.Log(". _dt.Rows[" + r + "][0]= " + _dt.Rows[r][0]);
#endif
if (Utility.isWordstart(pos))
{
decr = false;
result = new OrthographicResult();
result.Orthography = String.Empty;
result.Confi = 0f;
result.Level = String.Empty;
result.Phons = new List <string>();
#if DEBUG
logfile.Log(". . _dt.Rows[" + r + "][1]= " + _dt.Rows[r][1]);
logfile.Log(". . _dt.Rows[" + r + "][2]= " + _dt.Rows[r][2]);
logfile.Log(". . _dt.Rows[" + r + "][3]= " + _dt.Rows[r][3]);
#endif
result.Phons.Add(_dt.Rows[r][1] as String); // phon
result.Start = Decimal.Parse(_dt.Rows[r][2].ToString()); // start
result.phStops.Add(Decimal.Parse(_dt.Rows[r][3].ToString())); // stop - 1st phon
if (r != _dt.Rows.Count - 1)
{
decr = true;
pos = _dt.Rows[++r][0] as String;
while (!Utility.isWordstart(pos))
{
#if DEBUG
logfile.Log(". . . _dt.Rows[" + r + "][1]= " + _dt.Rows[r][1]);
logfile.Log(". . . _dt.Rows[" + r + "][3]= " + _dt.Rows[r][3]);
#endif
result.Phons.Add(_dt.Rows[r][1] as String); // phon - 2+
result.phStops.Add(Decimal.Parse(_dt.Rows[r][3].ToString())); // stop - 2+
if (r == _dt.Rows.Count - 1)
{
break;
}
pos = _dt.Rows[++r][0] as String;
}
}
#if DEBUG
logfile.Log(". . _dt.Rows[" + r + "][3]= " + _dt.Rows[r][3]);
#endif
result.Stop = Decimal.Parse(_dt.Rows[r][3].ToString()); // stop - word
_f._ars_alt.Add(result);
if (decr)
{
--r;
}
}
}
if (_f._ars_alt.Count != 0)
{
_f.AlternateData();
}
}
// ulong GetAudioStreamPositionSeconds(string pos)
// {
// ulong sec = UInt64.Parse(pos);
//
// sec /= 2uL; // bytes per sample (16-bit)
// sec /= 44100; // samples per second
//
// return sec;
// }
/// <summary>
/// Handles 'SpInProcRecoContext.Recognition' event. Fires as the final
/// hypothesis for a phrase. Each word will be added to a list of
/// 'OrthographicResult's for the phrase.
/// WARNING: This can fire 2+ on the same file-stream causing the engine
/// to drop/reset important variables like 'PhraseInfo.StartTime' and
/// 'word.AudioStreamOffset' and 'word.AudioTimeOffset'
/// TODO: a fact that is exceedingly annoying to try to compensate for.
/// </summary>
/// <param name="StreamNumber"></param>
/// <param name="StreamPosition"></param>
/// <param name="RecognitionType"></param>
/// <param name="Result"></param>
void rc_Recognition(int StreamNumber, object StreamPosition, SpeechRecognitionType RecognitionType, ISpeechRecoResult Result)
{
#if DEBUG
logfile.Log();
logfile.Log("rc_Recognition() #" + StreamNumber + " StreamPosition= " + StreamPosition + " _generato= " + _generato);
logfile.Log(". RecognitionType= " + RecognitionType); // <- standard.
logfile.Log(". _phoneConverter.LanguageId= " + _phoneConverter.LanguageId);
logfile.Log(". " + Result.PhraseInfo.GetText()); // (0, -1, true)
logfile.Log(". _offset = " + _offset);
logfile.Log(". PhraseInfo.AudioStreamPosition= " + Result.PhraseInfo.AudioStreamPosition);
// logfile.Log(". . sec= " + GetAudioStreamPositionSeconds(Result.PhraseInfo.AudioStreamPosition.ToString()));
logfile.Log(". PhraseInfo.AudioSizeBytes = " + Result.PhraseInfo.AudioSizeBytes);
logfile.Log(". PhraseInfo.StartTime = " + Result.PhraseInfo.StartTime);
logfile.Log(". PhraseInfo.AudioSizeTime = " + Result.PhraseInfo.AudioSizeTime);
logfile.Log(". Result.PhraseInfo.Rule.Name= " + Result.PhraseInfo.Rule.Name); // <- blank.
logfile.Log(". Result.PhraseInfo.Rule.Id= " + Result.PhraseInfo.Rule.Id);
logfile.Log(". Result.PhraseInfo.Rule.EngineConfidence= " + Result.PhraseInfo.Rule.EngineConfidence);
logfile.Log(". Result.PhraseInfo.Rule.Confidence= " + Result.PhraseInfo.Rule.Confidence);
logfile.Log(". wordcount= " + Result.PhraseInfo.Elements.Count);
#endif
List <OrthographicResult> ars = null;
switch (_generato)
{
case Generator.Dictati: ars = _ars_def; break;
case Generator.Dialogi: ars = _ars_enh; break;
}
foreach (ISpeechPhraseElement word in Result.PhraseInfo.Elements)
{
#if DEBUG
logfile.Log(". . word= " + word.DisplayText);
logfile.Log(". . LexicalForm= " + word.LexicalForm);
logfile.Log(". . DisplayAttributes= " + word.DisplayAttributes);
logfile.Log(". . EngineConfidence= " + word.EngineConfidence);
logfile.Log(". . ActualConfidence= " + word.ActualConfidence);
var ids = (ushort[])word.Pronunciation;
foreach (var id in ids)
{
logfile.Log(". . . PhoneId= " + id + " - " + _phoneConverter.IdToPhone(id));
}
logfile.Log(". . word.AudioStreamOffset= " + word.AudioStreamOffset);
logfile.Log(". . word.AudioSizeBytes = " + word.AudioSizeBytes);
logfile.Log(". . word.AudioTimeOffset = " + word.AudioTimeOffset);
logfile.Log(". . word.AudioSizeTime = " + word.AudioSizeTime);
#endif
var ar = new OrthographicResult();
ar.Orthography = word.DisplayText;
string phons = _phoneConverter.IdToPhone(word.Pronunciation); // NOTE: object is a ushort or ushort[]
ar.Phons = new List <string>(phons.Split(' '));
ar.Confi = word.EngineConfidence;
ar.Level = word.ActualConfidence.ToString().Replace("SEC", String.Empty).Replace("Confidence", String.Empty);
ar.Start = _offset + Utility.GarpstoSecs(word.AudioTimeOffset);
ar.Stop = _offset + Utility.GarpstoSecs(word.AudioTimeOffset + word.AudioSizeTime);
ars.Add(ar);
}
// NOTE: Recognition could be fired before the entire audiofile has
// completed, which means it's going to fire again but the AudioTimeOffsets
// will be completely borked obviously. So add this time-offset to any
// second or subsequent Recognition event that happens on this stream
_offset += Utility.GarpstoSecs(Result.PhraseInfo.AudioSizeTime); // TODO. is not accurate.
if (_text == String.Empty)
{
++Confidence_def_count;
Confidence_def += Result.PhraseInfo.Rule.EngineConfidence;
}
#if DEBUG
logfile.Log();
#endif
}
