// ################################# 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>
/// Initializes a new instance of the <see cref="AmplitudeSpectrogram"/> class.
/// </summary>
public AmplitudeSpectrogram(SpectrogramSettings config, WavReader wav)
{
this.Configuration = config;
this.Attributes = new SpectrogramAttributes();
double minDuration = 1.0;
if (wav.Time.TotalSeconds < minDuration)
{
LoggedConsole.WriteLine("Signal must at least {0} seconds long to produce a sonogram!", minDuration);
return;
}
//set attributes for the current recording and spectrogram type
this.Attributes.SampleRate = wav.SampleRate;
this.Attributes.Duration = wav.Time;
this.Attributes.NyquistFrequency = wav.SampleRate / 2;
this.Attributes.Duration = wav.Time;
this.Attributes.MaxAmplitude = wav.CalculateMaximumAmplitude();
this.Attributes.FrameDuration = TimeSpan.FromSeconds(this.Configuration.WindowSize / (double)wav.SampleRate);
var recording = new AudioRecording(wav);
var fftdata = DSP_Frames.ExtractEnvelopeAndFfts(
recording,
config.WindowSize,
config.WindowOverlap,
this.Configuration.WindowFunction);
// now recover required data
//epsilon is a signal dependent minimum amplitude value to prevent possible subsequent log of zero value.
this.Attributes.Epsilon = fftdata.Epsilon;
this.Attributes.WindowPower = fftdata.WindowPower;
this.Attributes.FrameCount = fftdata.FrameCount;
this.Data = fftdata.AmplitudeSpectrogram;
// IF REQUIRED CONVERT TO MEL SCALE
if (this.Configuration.DoMelScale)
{
// this mel scale conversion uses the "Greg integral" !
this.Data = MFCCStuff.MelFilterBank(this.Data, this.Configuration.MelBinCount, this.Attributes.NyquistFrequency, 0, this.Attributes.NyquistFrequency);
}
}
/// <summary>
/// Initializes a new instance of the <see cref="EnergySpectrogram"/> class.
/// Use this constructor when you have config and audio objects
/// </summary>
public EnergySpectrogram(SpectrogramSettings config, WavReader wav)
: this(new AmplitudeSpectrogram(config, wav))
{
}
