public static SignalFrame[] DivideSignalToFrames(double[] pSignal, int pSamplingRate, double pSignalLengthInMilliSeconds, double pFrameLengthinMilliSeconds)
{
int numberOfFrames = (int)Math.Ceiling(pSignalLengthInMilliSeconds / pFrameLengthinMilliSeconds);
//START FIX1
int frameSize = (int)(pSamplingRate * pFrameLengthinMilliSeconds / 1000.0);
//END FIX1
//Start FIX2
int remainingDataSize = pSignal.Length - frameSize * (numberOfFrames - 1);
int compensation = (int)(remainingDataSize / frameSize);
numberOfFrames += compensation;
remainingDataSize -= compensation * frameSize;
//End FIX2
//initialize frames.
SignalFrame[] frames = new SignalFrame[numberOfFrames];
for (int i = 0; i < numberOfFrames; i++)
{
//START: FIX1
frames[i] = new SignalFrame();
//END: FIX1
frames[i].Data = new double[frameSize];
}
//copy data from signal to frames.
int signalIndex = 0;
//START FIX1
for (int i = 0; i < numberOfFrames - 1; i++)
{
Array.Copy(pSignal, signalIndex, frames[i].Data, 0, frameSize);
signalIndex += frameSize;
}
Array.Copy(pSignal, signalIndex, frames[numberOfFrames - 1].Data, 0, remainingDataSize);
//END FIX1
return frames;
}
public static SignalFrame[] DivideSignalToFrames(double[] pSignal, int pSamplingRate, double pSignalLengthInMilliSeconds, double pFrameLengthinMilliSeconds)
{
int numberOfFrames = (int)Math.Ceiling(pSignalLengthInMilliSeconds / pFrameLengthinMilliSeconds);
//START FIX1
int frameSize = (int)(pSamplingRate * pFrameLengthinMilliSeconds / 1000.0);
//END FIX1
//Start FIX2
int remainingDataSize = pSignal.Length - frameSize * (numberOfFrames - 1);
int compensation = (int)(remainingDataSize / frameSize);
numberOfFrames += compensation;
remainingDataSize -= compensation * frameSize;
//End FIX2
//initialize frames.
SignalFrame[] frames = new SignalFrame[numberOfFrames];
for (int i = 0; i < numberOfFrames; i++)
{
//START: FIX1
frames[i] = new SignalFrame();
//END: FIX1
frames[i].Data = new double[frameSize];
}
//copy data from signal to frames.
int signalIndex = 0;
//START FIX1
for (int i = 0; i < numberOfFrames - 1; i++)
{
Array.Copy(pSignal, signalIndex, frames[i].Data, 0, frameSize);
signalIndex += frameSize;
}
Array.Copy(pSignal, signalIndex, frames[numberOfFrames - 1].Data, 0, remainingDataSize);
//END FIX1
return(frames);
}
public static SignalFrame[] DivideSignalToFrames(double[] pSignal, int pSamplingRate, double pSignalLengthInMilliSeconds, double pFrameLengthinMilliSeconds)
{
int numberOfFrames = (int)Math.Floor(pSignalLengthInMilliSeconds / pFrameLengthinMilliSeconds);
int frameSize = (int)(pSamplingRate * pFrameLengthinMilliSeconds / 100.0);
//initialize frames.
SignalFrame[] frames = new SignalFrame[numberOfFrames];
for (int i = 0; i < numberOfFrames; i++)
{
frames[i].Data = new double[frameSize];
}
//copy data from signal to frames.
int signalIndex = 0;
for (int i = 0; i < numberOfFrames; i++)
{
pSignal.CopyTo(frames[i].Data, signalIndex);
signalIndex += frameSize;
}
return(frames);
}
//Voice Activation Detection (VAD)
public static SignalFrame[] RemoveSilentSegments(SignalFrame[] pFrames)
{
double[] framesWeights = new double[pFrames.Length];
int frameIndex = 0;
foreach (SignalFrame frame in pFrames)
{
double squareMean = 0;
double avgZeroCrossing = 0;
for (int i = 0; i < frame.Data.Length - 1; i++)
{
//FIX1
squareMean += frame.Data[i] * frame.Data[i];
// avgZeroCrossing += Math.Abs(Math.Sign(frame.Data[i+1]) - Math.Sign(frame.Data[i])) / 2;
avgZeroCrossing += Math.Abs(Math.Abs(frame.Data[i + 1]) - Math.Abs(frame.Data[i])) / 2.0;
}
squareMean /= frame.Data.Length;
avgZeroCrossing /= frame.Data.Length;
framesWeights[frameIndex++] = squareMean * (1 - avgZeroCrossing) * 1000;
}
double avgWeights = mean(framesWeights);
double stdWeights = std(framesWeights);
double gamma = 0.2 * Math.Pow(stdWeights, -0.8);
double activationThreshold = avgWeights + gamma * stdWeights;
//threshold weights.
threshold(framesWeights, activationThreshold);
//smooth weights to remove short silences.
smooth(framesWeights);
//set anything more than 0 with 1.
threshold(framesWeights, 0);
int numberOfActiveFrames = (int)framesWeights.Sum();
SignalFrame[] activeFrames = new SignalFrame[numberOfActiveFrames];
int activeFramesIndex = 0;
for (int i = 0; i < pFrames.Length; i++)
{
if (framesWeights[i] == 1)
{
activeFrames[activeFramesIndex] = new SignalFrame();
activeFrames[activeFramesIndex].Data = new double[pFrames[i].Data.Length];
pFrames[i].Data.CopyTo(activeFrames[activeFramesIndex].Data, 0);
activeFramesIndex++;
}
}
return(activeFrames);
}
//Voice Activation Detection (VAD)
public static SignalFrame[] RemoveSilentSegments(SignalFrame[] pFrames)
{
double[] framesWeights = new double[pFrames.Length];
int frameIndex = 0;
foreach (SignalFrame frame in pFrames)
{
double squareMean = 0;
double avgZeroCrossing = 0;
for (int i = 0; i < frame.Data.Length - 1; i++)
{
//FIX1
squareMean += frame.Data[i] * frame.Data[i];
//avgZeroCrossing += Math.Abs(Math.Sign(frame.Data[i+1]) - Math.Sign(frame.Data[i])) / 2;
avgZeroCrossing += Math.Abs(Math.Abs(frame.Data[i + 1]) - Math.Abs(frame.Data[i])) / 2.0;
}
squareMean /= frame.Data.Length;
avgZeroCrossing /= frame.Data.Length;
framesWeights[frameIndex++] = squareMean * (1 - avgZeroCrossing) * 1000;
}
double avgWeights = mean(framesWeights);
double stdWeights = std(framesWeights);
double gamma = 0.2 * Math.Pow(stdWeights, -0.8);
double activationThreshold = avgWeights + gamma * stdWeights;
//threshold weights.
threshold(framesWeights, activationThreshold);
//smooth weights to remove short silences.
smooth(framesWeights);
//set anything more than 0 with 1.
threshold(framesWeights, 0);
int numberOfActiveFrames = (int)framesWeights.Sum();
SignalFrame[] activeFrames = new SignalFrame[numberOfActiveFrames];
int activeFramesIndex = 0;
for (int i = 0; i < pFrames.Length; i++)
{
if (framesWeights[i] == 1)
{
activeFrames[activeFramesIndex] = new SignalFrame();
activeFrames[activeFramesIndex].Data = new double[pFrames[i].Data.Length];
pFrames[i].Data.CopyTo(activeFrames[activeFramesIndex].Data, 0);
activeFramesIndex++;
}
}
return activeFrames;
}
