/// <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="BaseSonogram"/> class.
/// BASE CONSTRUCTOR
/// This constructor contains all steps required to prepare the amplitude spectrogram.
/// The third boolean parameter is simply a place-filler to ensure a different Constructor signature.
/// from the principle Constructor which follows.
/// </summary>
/// <param name="config">config file to use.</param>
/// <param name="wav">wav.</param>
/// <param name="dummy">filler boolean. Calculate in method.</param>
public BaseSonogram(SonogramConfig config, WavReader wav, bool dummy)
: this(config)
{
// As of 28 March 2017 drop capability to get sub-band of spectrogram because was not being used.
// can be recovered later if desired.
//bool doExtractSubband = this.SubBandMinHz > 0 || this.SubBandMaxHz < this.NyquistFrequency;
this.Duration = wav.Time;
double minDuration = 0.2;
if (this.Duration.TotalSeconds < minDuration)
{
LoggedConsole.WriteLine("Signal must at least {0} seconds long to produce a sonogram!", minDuration);
return;
}
//set config params to the current recording
this.SampleRate = wav.SampleRate;
this.Configuration.Duration = wav.Time;
this.Configuration.SampleRate = wav.SampleRate; //also set the Nyquist
this.MaxAmplitude = wav.CalculateMaximumAmplitude();
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.Configuration.epsilon = fftData.Epsilon;
this.Configuration.WindowPower = fftData.WindowPower;
this.FrameCount = fftData.FrameCount;
this.DecibelsPerFrame = fftData.FrameDecibels;
//init normalised signal energy array but do nothing with it. This has to be done from outside
this.DecibelsNormalised = new double[this.FrameCount];
this.Data = fftData.AmplitudeSpectrogram;
// ENERGY PER FRAME and NORMALISED dB PER FRAME AND SNR
// currently DoSnr = true by default
if (config.DoSnr)
{
// If the FractionOfHighEnergyFrames PRIOR to noise removal exceeds SNR.FractionalBoundForMode,
// then Lamel's noise removal algorithm may not work well.
if (fftData.FractionOfHighEnergyFrames > SNR.FractionalBoundForMode)
{
Log.WriteIfVerbose("\nWARNING ##############");
Log.WriteIfVerbose(
"\t############### BaseSonogram(): This is a high energy recording. Percent of high energy frames = {0:f0} > {1:f0}%",
fftData.FractionOfHighEnergyFrames * 100,
SNR.FractionalBoundForMode * 100);
Log.WriteIfVerbose("\t############### Noise reduction algorithm may not work well in this instance!\n");
}
//AUDIO SEGMENTATION/END POINT DETECTION - based on Lamel et al
// Setting segmentation/endpoint detection parameters is broken as of September 2014.
// The next line is a hack replacement
EndpointDetectionConfiguration.SetDefaultSegmentationConfig();
this.SigState = EndpointDetectionConfiguration.DetermineVocalisationEndpoints(this.DecibelsPerFrame, this.FrameStep);
}
/* AS OF 30 MARCH 2017, NO LONGER IMPLEMENT SUB-BAND THINGS, because not being used for years.
* // EXTRACT REQUIRED FREQUENCY BAND
* if (doExtractSubband)
* {
* this.Data = SpectrogramTools.ExtractFreqSubband(
* this.Data,
* this.subBandMinHz,
* this.subBandMaxHz,
* this.Configuration.DoMelScale,
* this.Configuration.FreqBinCount,
* this.FBinWidth);
* this.CalculateSubbandSNR(this.Data);
* }
*/
}
