private void WriteDebugImage(
AudioRecording recording,
DirectoryInfo outputDirectory,
BaseSonogram sonogram,
List <AcousticEvent> acousticEvents,
List <Plot> plots,
double[,] hits)
{
//DEBUG IMAGE this recogniser only. MUST set false for deployment.
bool displayDebugImage = MainEntry.InDEBUG;
if (displayDebugImage)
{
Image debugImage1 = SpectrogramTools.GetSonogramPlusCharts(sonogram, acousticEvents, plots, hits);
var debugPath1 = outputDirectory.Combine(FilenameHelpers.AnalysisResultName(Path.GetFileNameWithoutExtension(recording.BaseName), this.Identifier, "png", "DebugSpectrogram1"));
debugImage1.Save(debugPath1.FullName);
// save new image with longer frame
var sonoConfig2 = new SonogramConfig
{
SourceFName = recording.BaseName,
WindowSize = 1024,
WindowOverlap = 0,
//NoiseReductionType = NoiseReductionType.NONE,
NoiseReductionType = NoiseReductionType.Standard,
NoiseReductionParameter = 0.1,
};
BaseSonogram sonogram2 = new SpectrogramStandard(sonoConfig2, recording.WavReader);
var debugPath2 = outputDirectory.Combine(FilenameHelpers.AnalysisResultName(Path.GetFileNameWithoutExtension(recording.BaseName), this.Identifier, "png", "DebugSpectrogram2"));
Image debugImage2 = SpectrogramTools.GetSonogramPlusCharts(sonogram2, acousticEvents, plots, null);
debugImage2.Save(debugPath2.FullName);
}
}
/// <summary>
/// THis method can be modified if want to do something non-standard with the output spectrogram.
/// </summary>
internal static void SaveDebugSpectrogram(RecognizerResults results, Config genericConfig, DirectoryInfo outputDirectory, string baseName)
{
//var image = sonogram.GetImageFullyAnnotated("Test");
var image = SpectrogramTools.GetSonogramPlusCharts(results.Sonogram, results.Events, results.Plots, null);
image.Save(Path.Combine(outputDirectory.FullName, baseName + ".profile.png"));
}
/*
* /// <summary>
* /// Summarize your results. This method is invoked exactly once per original file.
* /// </summary>
* public override void SummariseResults(
* AnalysisSettings settings,
* FileSegment inputFileSegment,
* EventBase[] events,
* SummaryIndexBase[] indices,
* SpectralIndexBase[] spectralIndices,
* AnalysisResult2[] results)
* {
* // No operation - do nothing. Feel free to add your own logic.
* base.SummariseResults(settings, inputFileSegment, events, indices, spectralIndices, results);
* }
*/
/// <summary>
/// THis method can be modified if want to do something non-standard with the output spectrogram.
/// </summary>
public static string SaveDebugSpectrogram(RecognizerResults results, Config genericConfig, DirectoryInfo outputDirectory, string baseName)
{
var image3 = SpectrogramTools.GetSonogramPlusCharts(results.Sonogram, results.NewEvents, results.Plots, null);
var path = Path.Combine(outputDirectory.FullName, baseName + ".profile.png");
image3.Save(path);
return(path);
}
public void TestAnnotatedSonogramWithPlots()
{
// Make a decibel spectrogram
var actualDecibelSpectrogram = new SpectrogramStandard(this.sonoConfig, this.recording.WavReader);
// prepare normalisation bounds for three plots
double minDecibels = -100.0;
double maxDecibels = -50;
//double decibelThreshold = 12.5 dB above -100 dB;
var normThreshold = 0.25;
//plot 1
int minHz = 2000;
int maxHz = 3000;
var decibelArray = SNR.CalculateFreqBandAvIntensity(actualDecibelSpectrogram.Data, minHz, maxHz, actualDecibelSpectrogram.NyquistFrequency);
var normalisedIntensityArray = DataTools.NormaliseInZeroOne(decibelArray, minDecibels, maxDecibels);
var plot1 = new Plot("Intensity 2-3 kHz", normalisedIntensityArray, normThreshold);
//plot 2
minHz = 3000;
maxHz = 4000;
decibelArray = SNR.CalculateFreqBandAvIntensity(actualDecibelSpectrogram.Data, minHz, maxHz, actualDecibelSpectrogram.NyquistFrequency);
normalisedIntensityArray = DataTools.NormaliseInZeroOne(decibelArray, minDecibels, maxDecibels);
var plot2 = new Plot("Intensity 3-4 kHz", normalisedIntensityArray, normThreshold);
//plot 3
minHz = 4000;
maxHz = 5000;
decibelArray = SNR.CalculateFreqBandAvIntensity(actualDecibelSpectrogram.Data, minHz, maxHz, actualDecibelSpectrogram.NyquistFrequency);
normalisedIntensityArray = DataTools.NormaliseInZeroOne(decibelArray, minDecibels, maxDecibels);
var plot3 = new Plot("Intensity 4-5 kHz", normalisedIntensityArray, normThreshold);
// combine the plots
var plots = new List <Plot> {
plot1, plot2, plot3
};
// create three events
var startOffset = TimeSpan.Zero;
var events = new List <AcousticEvent>
{
new AcousticEvent(startOffset, 10.0, 10.0, 2000, 3000),
new AcousticEvent(startOffset, 25.0, 10.0, 3000, 4000),
new AcousticEvent(startOffset, 40.0, 10.0, 4000, 5000),
};
var image = SpectrogramTools.GetSonogramPlusCharts(actualDecibelSpectrogram, events, plots, null);
// create the image for visual confirmation
image.Save(Path.Combine(this.outputDirectory.FullName, this.recording.BaseName + ".png"));
Assert.AreEqual(1621, image.Width);
Assert.AreEqual(647, image.Height);
}
protected virtual Image DrawSonogram(
BaseSonogram sonogram,
double[,] hits,
List <Plot> scores,
List <AcousticEvent> predictedEvents,
double eventThreshold)
{
var image = SpectrogramTools.GetSonogramPlusCharts(sonogram, predictedEvents, scores, hits);
return(image);
}
/// <summary>
/// Do your analysis. This method is called once per segment (typically one-minute segments).
/// </summary>
public override RecognizerResults Recognize(AudioRecording recording, Config configuration, TimeSpan segmentStartOffset, Lazy <IndexCalculateResult[]> getSpectralIndexes, DirectoryInfo outputDirectory, int?imageWidth)
{
string speciesName = configuration[AnalysisKeys.SpeciesName] ?? "<no species>";
string abbreviatedSpeciesName = configuration[AnalysisKeys.AbbreviatedSpeciesName] ?? "<no.sp>";
const int frameSize = 256;
const double windowOverlap = 0.0;
double noiseReductionParameter = configuration.GetDoubleOrNull("SeverityOfNoiseRemoval") ?? 2.0;
int minHz = configuration.GetInt(AnalysisKeys.MinHz);
int maxHz = configuration.GetInt(AnalysisKeys.MaxHz);
// ignore oscillations below this threshold freq
int minOscilFreq = configuration.GetInt(AnalysisKeys.MinOscilFreq);
// ignore oscillations above this threshold freq
int maxOscilFreq = configuration.GetInt(AnalysisKeys.MaxOscilFreq);
// duration of DCT in seconds
//double dctDuration = (double)configuration[AnalysisKeys.DctDuration];
// minimum acceptable value of a DCT coefficient
double dctThreshold = configuration.GetDouble(AnalysisKeys.DctThreshold);
// min duration of event in seconds
double minDuration = configuration.GetDouble(AnalysisKeys.MinDuration);
// max duration of event in seconds
double maxDuration = configuration.GetDouble(AnalysisKeys.MaxDuration);
// min score for an acceptable event
double decibelThreshold = configuration.GetDouble(AnalysisKeys.DecibelThreshold);
// min score for an acceptable event
double eventThreshold = configuration.GetDouble(AnalysisKeys.EventThreshold);
if (recording.WavReader.SampleRate != 22050)
{
throw new InvalidOperationException("Requires a 22050Hz file");
}
// i: MAKE SONOGRAM
var sonoConfig = new SonogramConfig
{
SourceFName = recording.BaseName,
WindowSize = frameSize,
WindowOverlap = windowOverlap,
NoiseReductionType = NoiseReductionType.Standard,
NoiseReductionParameter = noiseReductionParameter,
};
var recordingDuration = recording.Duration;
int sr = recording.SampleRate;
double freqBinWidth = sr / (double)sonoConfig.WindowSize;
int minBin = (int)Math.Round(minHz / freqBinWidth) + 1;
int maxBin = (int)Math.Round(maxHz / freqBinWidth) + 1;
// duration of DCT in seconds - want it to be about 3X or 4X the expected maximum period
double framesPerSecond = freqBinWidth;
double minPeriod = 1 / (double)maxOscilFreq;
double maxPeriod = 1 / (double)minOscilFreq;
double dctDuration = 5 * maxPeriod;
// duration of DCT in frames
int dctLength = (int)Math.Round(framesPerSecond * dctDuration);
// set up the cosine coefficients
double[,] cosines = MFCCStuff.Cosines(dctLength, dctLength);
BaseSonogram sonogram = new SpectrogramStandard(sonoConfig, recording.WavReader);
int rowCount = sonogram.Data.GetLength(0);
double[] amplitudeArray = MatrixTools.GetRowAveragesOfSubmatrix(sonogram.Data, 0, minBin, rowCount - 1, maxBin);
// remove baseline from amplitude array
var highPassFilteredSignal = DspFilters.SubtractBaseline(amplitudeArray, 7);
// remove hi freq content from amplitude array
var lowPassFilteredSignal = DataTools.filterMovingAverageOdd(amplitudeArray, 11);
var dctScores = new double[highPassFilteredSignal.Length];
const int step = 2;
for (int i = dctLength; i < highPassFilteredSignal.Length - dctLength; i += step)
{
if (highPassFilteredSignal[i] < decibelThreshold)
{
continue;
}
double[] subArray = DataTools.Subarray(highPassFilteredSignal, i, dctLength);
// Look for oscillations in the highPassFilteredSignal
Oscillations2014.GetOscillationUsingDct(subArray, framesPerSecond, cosines, out var oscilFreq, out var period, out var intensity);
bool periodWithinBounds = period > minPeriod && period < maxPeriod;
if (!periodWithinBounds)
{
continue;
}
if (intensity < dctThreshold)
{
continue;
}
//lay down score for sample length
for (int j = 0; j < dctLength; j++)
{
if (dctScores[i + j] < intensity && lowPassFilteredSignal[i + j] > decibelThreshold)
{
dctScores[i + j] = intensity;
}
}
}
//iii: CONVERT decibel sum-diff SCORES TO ACOUSTIC EVENTS
var acousticEvents = AcousticEvent.ConvertScoreArray2Events(
dctScores,
minHz,
maxHz,
sonogram.FramesPerSecond,
freqBinWidth,
eventThreshold,
minDuration,
maxDuration,
segmentStartOffset);
// ######################################################################
acousticEvents.ForEach(ae =>
{
ae.SpeciesName = speciesName;
ae.SegmentDurationSeconds = recordingDuration.TotalSeconds;
ae.SegmentStartSeconds = segmentStartOffset.TotalSeconds;
ae.Name = abbreviatedSpeciesName;
});
var plot = new Plot(this.DisplayName, dctScores, eventThreshold);
var plots = new List <Plot> {
plot
};
// DEBUG IMAGE this recognizer only. MUST set false for deployment.
bool displayDebugImage = MainEntry.InDEBUG;
if (displayDebugImage)
{
// display a variety of debug score arrays
DataTools.Normalise(amplitudeArray, decibelThreshold, out var normalisedScores, out var normalisedThreshold);
var ampltdPlot = new Plot("amplitude", normalisedScores, normalisedThreshold);
DataTools.Normalise(highPassFilteredSignal, decibelThreshold, out normalisedScores, out normalisedThreshold);
var demeanedPlot = new Plot("Hi Pass", normalisedScores, normalisedThreshold);
DataTools.Normalise(lowPassFilteredSignal, decibelThreshold, out normalisedScores, out normalisedThreshold);
var lowPassPlot = new Plot("Low Pass", normalisedScores, normalisedThreshold);
var debugPlots = new List <Plot> {
ampltdPlot, lowPassPlot, demeanedPlot, plot
};
Image debugImage = SpectrogramTools.GetSonogramPlusCharts(sonogram, acousticEvents, debugPlots, null);
var debugPath = outputDirectory.Combine(FilenameHelpers.AnalysisResultName(Path.GetFileNameWithoutExtension(recording.BaseName), this.Identifier, "png", "DebugSpectrogram"));
debugImage.Save(debugPath.FullName);
}
return(new RecognizerResults()
{
Sonogram = sonogram,
Hits = null,
Plots = plots,
Events = acousticEvents,
});
}