public List <SoundFeature> GetLPCAndCepstralSeries(string lpcNamePrefix, int lpcOrder, string cepstralNamePrefix, int cepstralOrder)
{
List <SoundFeature> lpcSoundFeatureList = new List <SoundFeature>();
for (int ii = 0; ii < lpcOrder; ii++)
{
SoundFeature soundFeature = new SoundFeature();
soundFeature.Name = lpcNamePrefix + (ii + 1).ToString(); // LPCs enumerated from 1.
soundFeature.SetSize(frameList.Count);
lpcSoundFeatureList.Add(soundFeature);
}
List <SoundFeature> cepstralSoundFeatureList = new List <SoundFeature>();
for (int ii = 0; ii <= cepstralOrder; ii++)
{
SoundFeature soundFeature = new SoundFeature();
soundFeature.Name = cepstralNamePrefix + ii.ToString();
soundFeature.SetSize(frameList.Count);
cepstralSoundFeatureList.Add(soundFeature);
}
List <double> lpcCoefficients = null;
for (int iFrame = 0; iFrame < frameList.Count; iFrame++)
{
WAVSound frame = frameList[iFrame];
lpcCoefficients = frame.ComputeLPCCoefficients(lpcOrder);
for (int ii = 0; ii < lpcOrder; ii++)
{
lpcSoundFeatureList[ii].ValueList[iFrame] = lpcCoefficients[ii];
}
List <double> cepstralCoefficients = frame.ComputeCepstralCoefficients(lpcCoefficients, cepstralOrder);
for (int ii = 0; ii <= cepstralOrder; ii++)
{
cepstralSoundFeatureList[ii].ValueList[iFrame] = cepstralCoefficients[ii];
}
}
//
// The first cepstral coefficient (c0) is equal to the (non-normalized) autocorrelation,
// and can usually be removed in speech recognition (the autocorrelation is typically computed
// elsewhere, as a separate feature).
//
// Moreover the second cepstral coefficient (c1) is identical to the first LPC coefficient,
// and so can also be removed.
//
cepstralSoundFeatureList.RemoveAt(0); // Remove c0
cepstralSoundFeatureList.RemoveAt(0); // Remove c1
List <SoundFeature> soundFeatureList = new List <SoundFeature>();
soundFeatureList.AddRange(lpcSoundFeatureList);
soundFeatureList.AddRange(cepstralSoundFeatureList);
return(soundFeatureList);
}
private void ShowTestSoundFeature(SoundFeature soundFeature)
{
featureComparisonPlotPanel.Clear();
featureComparisonPlotPanel.VerticalAutoRange = true;
DataSeries dataSeries = new DataSeries();
dataSeries.Generate(soundFeature.Name, soundFeature.TimeList, soundFeature.ValueList);
dataSeries.SetPointVisibilityState(true);
dataSeries.SetPointConnectionState(true);
dataSeries.SetLineColor(Color.Cyan);
dataSeries.SetPointColor(Color.Blue);
featureComparisonPlotPanel.AddDataSeries(dataSeries);
}
//
public SoundFeature GetRelativeNumberOfZeroCrossingsSeries(string name)
{
SoundFeature rnzcFeature = new SoundFeature();
rnzcFeature.Name = name;
rnzcFeature.SetSize(frameList.Count);
for (int iFrame = 0; iFrame < frameList.Count; iFrame++)
{
WAVSound frame = frameList[iFrame];
double relativeNumberOfZeroCrossings = frame.GetRelativeNumberOfZeroCrossings(0); // 0 = channel index (again assuming mono).
rnzcFeature.ValueList[iFrame] = relativeNumberOfZeroCrossings;
}
return(rnzcFeature);
}
public IWRRecognitionResult RecognizeSingle(WAVSound sound)
{
// Compute the features of the current sound
sound.SubtractMean();
double startTime = sound.GetFirstTimeAboveThreshold(0, soundExtractionMovingAverageLength,
soundExtractionThreshold);
double endTime = sound.GetLastTimeAboveThreshold(0, soundExtractionMovingAverageLength,
soundExtractionThreshold);
WAVSound extractedInstance = sound.Extract(startTime, endTime);
if (extractedInstance == null)
{
return(null);
} // 20170114
extractedInstance.PreEmphasize(preEmphasisThresholdFrequency);
WAVFrameSet frameSet = new WAVFrameSet(extractedInstance, frameDuration, frameShift);
frameSet.ApplyHammingWindows(alpha);
SoundFeatureSet soundFeatureSet = new SoundFeatureSet();
List <SoundFeature> autoCorrelationFeatureList = frameSet.GetAutoCorrelationSeries("AutoCorrelation", autoCorrelationOrder);
soundFeatureSet.FeatureList.AddRange(autoCorrelationFeatureList);
List <SoundFeature> lpcAndCepstralFeatureList = frameSet.GetLPCAndCepstralSeries("LPC", lpcOrder, "Cepstral", cepstralOrder);
soundFeatureSet.FeatureList.AddRange(lpcAndCepstralFeatureList);
SoundFeature relativeNumberOfZeroCrossingsFeature = frameSet.GetRelativeNumberOfZeroCrossingsSeries("RNZC");
soundFeatureSet.FeatureList.Add(relativeNumberOfZeroCrossingsFeature);
soundFeatureSet.SetNormalizedTime();
soundFeatureSet.Interpolate(numberOfValuesPerFeature);
IWRRecognitionResult recognitionResult = new IWRRecognitionResult();
recognitionResult.SoundFeatureSet = soundFeatureSet;
if (averageSoundFeatureSetList != null)
{
foreach (SoundFeatureSet averageSoundFeatureSet in averageSoundFeatureSetList)
{
double deviation = SoundFeatureSet.GetDeviation(averageSoundFeatureSet, soundFeatureSet, weightList);
string soundName = averageSoundFeatureSet.Information;
recognitionResult.DeviationList.Add(new Tuple <string, double>(soundName, deviation));
}
recognitionResult.DeviationList.Sort((a, b) => a.Item2.CompareTo(b.Item2));
}
return(recognitionResult);
}
public List <SoundFeature> GetAutoCorrelationSeries(string namePrefix, int autoCorrelationOrder)
{
List <SoundFeature> soundFeatureList = new List <SoundFeature>();
for (int ii = 0; ii < autoCorrelationOrder; ii++)
{
SoundFeature soundFeature = new SoundFeature();
soundFeature.Name = namePrefix + (ii + 1).ToString(); // No reason to compute order 0 (=1).
soundFeature.SetSize(frameList.Count);
soundFeatureList.Add(soundFeature);
}
for (int iFrame = 0; iFrame < frameList.Count; iFrame++)
{
WAVSound frame = frameList[iFrame];
List <double> autoCorrelationCoefficients = frame.ComputeNormalizedAutoCorrelationCoefficients(1, autoCorrelationOrder);
for (int ii = 0; ii < autoCorrelationOrder; ii++)
{
soundFeatureList[ii].ValueList[iFrame] = autoCorrelationCoefficients[ii];
}
}
return(soundFeatureList);
}
private void ShowComparisonSoundFeature(SoundFeature comparisonSoundFeature)
{
DataSeries dataSeries = new DataSeries();
dataSeries.Generate(comparisonSoundFeature.Name, comparisonSoundFeature.TimeList, comparisonSoundFeature.ValueList);
dataSeries.SetPointVisibilityState(true);
dataSeries.SetPointConnectionState(true);
dataSeries.SetLineColor(Color.Black);
dataSeries.SetPointColor(Color.Red);
List <double> verticalErrorBarList = new List <double>();
for (int ii = 0; ii < comparisonSoundFeature.VarianceList.Count; ii++)
{
double standardDeviation = Math.Sqrt(comparisonSoundFeature.VarianceList[ii]);
verticalErrorBarList.Add(standardDeviation);
}
dataSeries.AddSymmetricVerticalErrorBars(verticalErrorBarList);
dataSeries.SetVerticalErrorBarSerifVisibilityState(true);
dataSeries.SetErrorBarRelativeSerifLength(0.002);
dataSeries.SetVerticalErrorBarsVisibilityState(true);
featureComparisonPlotPanel.AddDataSeries(dataSeries);
}
public void AppendSound(string name, List <WAVSound> instanceList)
{
List <SoundFeatureSet> soundFeatureSetList = new List <SoundFeatureSet>();
// Compute the sound feature set for each instance
foreach (WAVSound soundInstance in instanceList)
{
soundInstance.SubtractMean();
double startTime = soundInstance.GetFirstTimeAboveThreshold(0, soundExtractionMovingAverageLength,
soundExtractionThreshold);
double endTime = soundInstance.GetLastTimeAboveThreshold(0, soundExtractionMovingAverageLength,
soundExtractionThreshold);
WAVSound extractedInstance = soundInstance.Extract(startTime, endTime);
extractedInstance.PreEmphasize(preEmphasisThresholdFrequency);
WAVFrameSet frameSet = new WAVFrameSet(extractedInstance, frameDuration, frameShift);
frameSet.ApplyHammingWindows(alpha);
SoundFeatureSet soundFeatureSet = new SoundFeatureSet();
List <SoundFeature> autoCorrelationFeatureList = frameSet.GetAutoCorrelationSeries("AutoCorrelation", autoCorrelationOrder);
soundFeatureSet.FeatureList.AddRange(autoCorrelationFeatureList);
List <SoundFeature> lpcAndCepstralFeatureList = frameSet.GetLPCAndCepstralSeries("LPC", lpcOrder, "Cepstral", cepstralOrder);
soundFeatureSet.FeatureList.AddRange(lpcAndCepstralFeatureList);
SoundFeature relativeNumberOfZeroCrossingsFeature = frameSet.GetRelativeNumberOfZeroCrossingsSeries("RNZC");
soundFeatureSet.FeatureList.Add(relativeNumberOfZeroCrossingsFeature);
soundFeatureSet.SetNormalizedTime();
soundFeatureSet.Interpolate(numberOfValuesPerFeature);
soundFeatureSetList.Add(soundFeatureSet);
}
SoundFeatureSet averageSoundFeatureSet = SoundFeatureSet.GenerateAverage(soundFeatureSetList);
averageSoundFeatureSet.Information = name; // The name of the stored sound.
if (averageSoundFeatureSetList == null)
{
averageSoundFeatureSetList = new List <SoundFeatureSet>();
}
averageSoundFeatureSetList.Add(averageSoundFeatureSet);
averageSoundFeatureSetList.Sort((a, b) => a.Information.CompareTo(b.Information)); // Perhaps remove?
OnAvailableSoundsChanged();
}
private void ShowTestSoundAndComparisonFeatures()
{
featureComparisonPlotPanel.Clear();
int selectedFeatureIndex = featureComparisonComboBox.SelectedIndex;
// Plot the feature values from the test sound:
SoundFeature soundFeature = recognitionResult.SoundFeatureSet.FeatureList[selectedFeatureIndex];
ShowTestSoundFeature(soundFeature);
// A bit ugly, but OK...
if (deviationListBox.SelectedIndex >= 0)
{
string selectedComparisonSound =
deviationListBox.Items[deviationListBox.SelectedIndex].ToString().Split(new char[] { ' ' }, StringSplitOptions.RemoveEmptyEntries)[0];
SoundFeatureSet comparisonSoundFeatureSet = recognizer.AverageSoundFeatureSetList.Find(s => s.Information == selectedComparisonSound);
if (comparisonSoundFeatureSet != null) // Should always be the case...
{
SoundFeature comparisonSoundFeature = comparisonSoundFeatureSet.FeatureList[selectedFeatureIndex];
ShowComparisonSoundFeature(comparisonSoundFeature);
}
}
}
private void AppendFeatureSeries(int selectedSoundIndex, int selectedFeatureIndex)
{
SoundFeature averageSoundFeature = recognizer.AverageSoundFeatureSetList[selectedSoundIndex].FeatureList[selectedFeatureIndex];
string featureName = averageSoundFeature.Name;
DataSeries dataSeries = new DataSeries();
dataSeries.Generate(featureName, averageSoundFeature.TimeList, averageSoundFeature.ValueList);
dataSeries.SetPointVisibilityState(true);
dataSeries.SetPointConnectionState(true);
dataSeries.SetLineColor(Color.Black);
dataSeries.SetPointColor(Color.Red);
List <double> verticalErrorBarList = new List <double>();
for (int ii = 0; ii < averageSoundFeature.VarianceList.Count; ii++)
{
double standardDeviation = Math.Sqrt(averageSoundFeature.VarianceList[ii]);
verticalErrorBarList.Add(standardDeviation);
}
dataSeries.AddSymmetricVerticalErrorBars(verticalErrorBarList);
dataSeries.SetVerticalErrorBarSerifVisibilityState(true);
dataSeries.SetErrorBarRelativeSerifLength(0.002);
featurePlotPanel.AddDataSeries(dataSeries);
}
