public static Image FrameSonogram(Image image, Image titleBar, TimeSpan minuteOffset, TimeSpan xAxisTicInterval, TimeSpan xAxisPixelDuration,
TimeSpan labelInterval, int nyquist, int herzInterval)
{
double secondsDuration = xAxisPixelDuration.TotalSeconds * image.Width;
var fullDuration = TimeSpan.FromSeconds(secondsDuration);
// init frequency scale
int frameSize = image.Height;
var freqScale = new FrequencyScale(nyquist, frameSize, herzInterval);
SpectrogramTools.DrawGridLinesOnImage((Bitmap)image, minuteOffset, fullDuration, xAxisTicInterval, freqScale);
int imageWidth = image.Width;
int trackHeight = 20;
int imageHt = image.Height + trackHeight + trackHeight + trackHeight;
var timeBmp = DrawTimeTrack(minuteOffset, xAxisPixelDuration, xAxisTicInterval, labelInterval, imageWidth, trackHeight, "Seconds");
var compositeBmp = new Bitmap(imageWidth, imageHt); //get canvas for entire image
var gr = Graphics.FromImage(compositeBmp);
gr.Clear(Color.Black);
int offset = 0;
gr.DrawImage(titleBar, 0, offset); //draw in the top time scale
offset += timeBmp.Height;
gr.DrawImage(timeBmp, 0, offset); //draw
offset += titleBar.Height;
gr.DrawImage(image, 0, offset); //draw
offset += image.Height;
gr.DrawImage(timeBmp, 0, offset); //draw
return(compositeBmp);
}
public SpectrogramCepstral(AmplitudeSonogram sg, int minHz, int maxHz)
: this(sg)
{
this.DecibelsPerFrame = sg.DecibelsPerFrame;
this.DecibelsNormalised = sg.DecibelsNormalised;
this.Duration = sg.Duration;
this.FrameCount = sg.FrameCount;
this.DecibelReference = sg.DecibelReference;
this.MaxAmplitude = sg.MaxAmplitude;
this.SampleRate = sg.SampleRate;
this.SigState = sg.SigState;
this.SnrData = sg.SnrData;
// subband highlighting no longer available
//this.subBandMinHz = minHz;
//this.subBandMaxHz = maxHz;
//double[] noise_subband = BaseSonogram.ExtractModalNoiseSubband(this.SnrData.ModalNoiseProfile, minHz, maxHz, sg.doMelScale,
// sonogram.Configuration.FreqBinCount, sonogram.FBinWidth);
this.Data = SpectrogramTools.ExtractFreqSubband(sg.Data, minHz, maxHz,
this.Configuration.DoMelScale, sg.Configuration.FreqBinCount, sg.FBinWidth);
// NO LONGER DO THIS >>>> CalculateSubbandSNR(this.Data);
this.Make(this.Data); //converts amplitude matrix to cepstral sonogram
}
/// <summary>
/// This method assumes that the height of the passed sonogram image is half of the original frame size.
/// This assumption allows the frequency scale grid lines to be placed at the correct intervals.
/// </summary>
public static Image FrameSonogram(
Image sonogramImage,
Image titleBar,
TimeSpan minuteOffset,
TimeSpan xAxisTicInterval,
TimeSpan xAxisPixelDuration,
TimeSpan labelInterval,
int nyquist,
int hertzInterval)
{
double secondsDuration = xAxisPixelDuration.TotalSeconds * sonogramImage.Width;
var fullDuration = TimeSpan.FromSeconds(secondsDuration);
// init frequency scale
int frameSize = sonogramImage.Height * 2;
var freqScale = new FrequencyScale(nyquist, frameSize, hertzInterval);
SpectrogramTools.DrawGridLinesOnImage((Bitmap)sonogramImage, minuteOffset, fullDuration, xAxisTicInterval, freqScale);
int imageWidth = sonogramImage.Width;
var timeBmp = ImageTrack.DrawShortTimeTrack(minuteOffset, xAxisPixelDuration, xAxisTicInterval, labelInterval, imageWidth, "Seconds");
Image[] imageArray = { titleBar, timeBmp, sonogramImage, timeBmp };
return(ImageTools.CombineImagesVertically(imageArray));
}
/// <summary>
/// Use this method to average a decibel spectrogram
/// </summary>
public static double[] CalculateAvgDecibelSpectrumFromSpectrogram(double[,] spectrogram)
{
int frameCount = spectrogram.GetLength(0);
int freqBinCount = spectrogram.GetLength(1);
double[] avgSpectrum = new double[freqBinCount];
for (int j = 0; j < freqBinCount; j++)
{
var freqBin = MatrixTools.GetColumn(spectrogram, j);
double av = SpectrogramTools.AverageAnArrayOfDecibelValues(freqBin);
avgSpectrum[j] = av;
}
return(avgSpectrum);
}
/// <summary>
/// The data passed to this method must be the Spectral sonogram.
/// </summary>
public static Tuple <double[, ], double[]> GetCepstrogram(double[,] data, int minHz, int maxHz,
int freqBinCount, double freqBinWidth, bool doMelScale, int ccCount)
{
ImageTools.DrawMatrix(data, @"C:\SensorNetworks\Output\MFCC_LewinsRail\tempImage1.jpg", false);
double[,] m = SpectrogramTools.ExtractFreqSubband(data, minHz, maxHz, doMelScale, freqBinCount, freqBinWidth);
ImageTools.DrawMatrix(m, @"C:\SensorNetworks\Output\MFCC_LewinsRail\tempImage2.jpg", false);
//DO NOT DO NOISE REDUCTION BECAUSE ALREADY DONE
//double[] modalNoise = SNR.CalculateModalNoise(m, 7); //calculate modal noise profile and smooth
//m = SNR.NoiseReduce_Standard(m, modalNoise);
//m = SNR.NoiseReduce_FixedRange(m, this.Configuration.DynamicRange);
m = MFCCStuff.Cepstra(m, ccCount);
m = DataTools.normalise(m);
ImageTools.DrawMatrix(m, @"C:\SensorNetworks\Output\MFCC_LewinsRail\tempImage3.jpg", false);
return(Tuple.Create(m, (double[])null));
}
// ################################# STATIC METHODS BELOW HERE ###############################
public static Image DrawSpectrogramAnnotated(double[,] data, SpectrogramSettings config, SpectrogramAttributes attributes)
{
// normalise the data between 0 and 95th percentiles
int binCount = 100;
double min;
double max;
DataTools.MinMax(data, out min, out max);
double binWidth = (max - min) / binCount;
var histogram = Histogram.Histo(data, binCount, min, max, binWidth);
int percentile = 95;
int binId = Histogram.GetPercentileBin(histogram, percentile);
double upperBound = min + (binId * percentile);
var normedMatrix = MatrixTools.NormaliseInZeroOne(data, min, upperBound);
/*
* int minPercentile = 5;
* int minBinId = Histogram.GetPercentileBin(histogram, minPercentile);
* double lowerBound = min + (minBinId * minPercentile);
* int maxPercentile = 95;
* int maxBinId = Histogram.GetPercentileBin(histogram, maxPercentile);
* double upperBound = min + (maxBinId * maxPercentile);
* var normedMatrix = MatrixTools.NormaliseInZeroOne(data, lowerBound, upperBound);
*/
int nyquist = attributes.NyquistFrequency;
int frameSize = config.WindowSize;
// assuming linear frequency scale
int finalBinCount = frameSize / 2;
var scaleType = FreqScaleType.Linear;
// if doing mel scale then
if (config.DoMelScale)
{
finalBinCount = 256; //128; //512; // 256; // 100; // 40; // 200; //
scaleType = FreqScaleType.Mel;
}
var freqScale = new FrequencyScale(scaleType, nyquist, frameSize, finalBinCount, hertzGridInterval: 1000);
var image = SpectrogramTools.GetImage(normedMatrix, nyquist, config.DoMelScale);
var annotatedImage = SpectrogramTools.GetImageFullyAnnotated(image, config.SourceFileName + ": " + scaleType.ToString(), freqScale.GridLineLocations, attributes.Duration);
return(annotatedImage);
}
/// <summary>
/// Returns a Spectrogram and Cepstrogram from the passed recording. These are NOT noise reduced.
/// however, tuple also returns the modal noise and sub-band modal noise.
/// </summary>
public static Tuple <SpectrogramStandard, SpectrogramCepstral, double[], double[]> GetAllSonograms(AudioRecording recording, SonogramConfig sonoConfig, int minHz, int maxHz)
{
int sr = recording.SampleRate;
bool doMelScale = sonoConfig.DoMelScale;
int ccCount = sonoConfig.mfccConfig.CcCount;
bool includeDelta = sonoConfig.mfccConfig.IncludeDelta;
bool includeDoubleDelta = sonoConfig.mfccConfig.IncludeDoubleDelta;
sonoConfig.SourceFName = recording.BaseName;
var basegram = new AmplitudeSonogram(sonoConfig, recording.WavReader);
var sonogram = new SpectrogramStandard(basegram); //spectrogram has dim[N,257]
Log.WriteLine("Signal: Duration={0}, Sample Rate={1}", sonogram.Duration, sr);
Log.WriteLine(
$"Frames: Size={0}, Count={1}, Duration={2:f1}ms, Overlap={5:f0}%, Offset={3:f1}ms, Frames/s={4:f1}",
sonogram.Configuration.WindowSize,
sonogram.FrameCount,
sonogram.FrameDuration * 1000,
sonogram.FrameStep * 1000,
sonogram.FramesPerSecond,
sonoConfig.WindowOverlap * 100);
int binCount = (int)(maxHz / sonogram.FBinWidth) - (int)(minHz / sonogram.FBinWidth) + 1;
Log.WriteLine("Freqs : {0} Hz - {1} Hz. (Freq bin count = {2})", minHz, maxHz, binCount);
Log.WriteLine("MFCCs : doMelScale=" + doMelScale + "; ccCount=" + ccCount + "; includeDelta=" + includeDelta + "; includeDoubleDelta=" + includeDoubleDelta);
//CALCULATE MODAL NOISE PROFILE - USER MAY REQUIRE IT FOR NOISE REDUCTION
double[] modalNoise = sonogram.SnrData.ModalNoiseProfile;
//extract sub-band modal noise profile
double[] noiseSubband = SpectrogramTools.ExtractModalNoiseSubband(modalNoise, minHz, maxHz, doMelScale, sonogram.NyquistFrequency, sonogram.FBinWidth);
// CALCULATE CEPSTRO-GRAM. //cepstrogram has dim[N,13]
var cepstrogram = new SpectrogramCepstral(basegram, minHz, maxHz);
var tuple = Tuple.Create(sonogram, cepstrogram, modalNoise, noiseSubband);
return(tuple);
}
