public virtual SpeechType GetFrameSpeechType(WAVSound frame)
// , int channel, double lowPassCutoffFrequency, double lowPassRatioThreshold,
// double energyThreshold, double silenceThreshold)
{
double zeroCrossingRate = frame.GetRelativeNumberOfZeroCrossings(channel);
double averageEnergy = frame.GetAverageEnergy(channel);
if (averageEnergy < silenceThreshold)
{
return(SpeechType.Silence);
}
else
{
WAVSound lowPassFilteredFrame = frame.Copy();
lowPassFilteredFrame.LowPassFilter(lowPassCutoffFrequency);
double lowPassFilteredAverageEnergy = lowPassFilteredFrame.GetAverageEnergy(channel);
double energyRatio = lowPassFilteredAverageEnergy / averageEnergy;
if ((energyRatio > lowPassRatioThreshold) && (averageEnergy > energyThreshold) && (zeroCrossingRate < zeroCrossingRateThreshold))
{
return(SpeechType.Voiced);
}
else if (zeroCrossingRate < zeroCrossingRateThreshold)
{
return(SpeechType.Voiced);
}
else
{
return(SpeechType.Unvoiced);
}
}
}
//
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);
}
