private List <Frame> SplitIntoFrames(double[] speechSignal, int frameWidth, int frameShift)
{
int frameCount = GetFramesCount(speechSignal.Length, frameWidth, frameShift);
var frames = new List <Frame>();
for (int i = 0; i < frameCount; i++)
{
int frameStart = i * frameShift;
int frameEnd = GetFrameEnd(i, frameWidth, frameShift, speechSignal.Length);
var frameBounds = new SpeechBounds(frameStart, frameEnd);
double[] frameSingal = new double[frameBounds.Width];
Array.Copy(speechSignal, frameStart, frameSingal, 0, frameSingal.Length);
var frame = new Frame(frameSingal, frameBounds);
frames.Add(frame);
}
return(frames);
}
public SignalContainer GetSpeech(SignalContainer rawSignal)
{
SignalContainer singal = rawSignal.Copy();
singal.FrameSize = 100; //ToDo! Constant change
singal.FrameShift = singal.FrameSize;
List <double> framesEnergy = singal.GetFrames().Select(frame => SignalHelper.GetAverageEnergy(frame.Signal)).ToList();
double averageEnergy = framesEnergy.Average();
double[] flattenedFramesEnergy = ApplyMeanFilter(framesEnergy.ToArray(), 3);
SpeechBounds speechFramesBounds = GetSpeechBounds(flattenedFramesEnergy, averageEnergy);
var speechFrames = singal.GetFrames().Skip(speechFramesBounds.Start).Take(speechFramesBounds.Width);
double[] speechSignal = speechFrames.SelectMany(frame => frame.Signal).ToArray();
return(new SignalContainer(speechSignal)
{
FrameSize = rawSignal.FrameSize,
FrameShift = rawSignal.FrameShift
});
}
public FrameFeatures(SpeechBounds frameBounds, double[] features)
{
FrameBounds = frameBounds;
vFeatures = new double[features.Length];
features.CopyTo(vFeatures, 0);
}
public Frame(double[] signal, SpeechBounds bounds)
{
Signal = new double[signal.Length];
signal.CopyTo(Signal, 0);
Bounds = bounds;
}
