/// <summary>
/// A one-frame track sounds like a click.
/// A click is a sharp onset broadband sound of brief duration. Geometrically it is similar to a vertical whistle.
/// THis method averages dB log values incorrectly but it is faster than doing many log conversions.
/// This method is used to find acoustic events and is accurate enough for the purpose.
/// </summary>
public static (List <EventCommon> Events, double[] Intensity) GetOneFrameTracks(
SpectrogramStandard sonogram,
OneframeTrackParameters parameters,
TimeSpan segmentStartOffset)
{
var sonogramData = sonogram.Data;
int frameCount = sonogramData.GetLength(0);
int binCount = sonogramData.GetLength(1);
var frameStep = sonogram.FrameStep;
int nyquist = sonogram.NyquistFrequency;
double binWidth = nyquist / (double)binCount;
int minBin = (int)Math.Round(parameters.MinHertz.Value / binWidth);
int maxBin = (int)Math.Round(parameters.MaxHertz.Value / binWidth);
var decibelThreshold = parameters.DecibelThreshold.Value;
var minBandwidthHertz = parameters.MinBandwidthHertz.Value;
var maxBandwidthHertz = parameters.MaxBandwidthHertz.Value;
var converter = new UnitConverters(
segmentStartOffset: segmentStartOffset.TotalSeconds,
sampleRate: sonogram.SampleRate,
frameSize: sonogram.Configuration.WindowSize,
frameOverlap: sonogram.Configuration.WindowOverlap);
// Find all frame peaks and place in peaks matrix
// avoid row edge effects.
var peaks = new double[frameCount, binCount];
// for all time frames except 1st and last allowing for edge effects.
for (int t = 1; t < frameCount - 1; t++)
{
// buffer zone around click is one frame wide.
// for all frequency bins except top and bottom in this time frame
for (int bin = minBin; bin < maxBin; bin++)
{
if (sonogramData[t, bin] < decibelThreshold)
{
continue;
}
// THis is where the profile of a click is defined
// A click requires sudden onset, with maximum amplitude followed by decay.
bool isClickPeak = sonogramData[t - 1, bin] < decibelThreshold && sonogramData[t, bin] > sonogramData[t + 1, bin];
if (isClickPeak)
{
peaks[t, bin] = sonogramData[t, bin];
}
}
}
//NOTE: the Peaks matrix is same size as the sonogram.
var tracks = GetOneFrameTracks(peaks, minBin, maxBin, minBandwidthHertz, maxBandwidthHertz, decibelThreshold, converter);
// initialise tracks as events and get the combined intensity array.
var events = new List <EventCommon>();
var temporalIntensityArray = new double[frameCount];
var maxScore = decibelThreshold * 5;
foreach (var track in tracks)
{
var ae = new ClickEvent(track, maxScore)
{
SegmentStartSeconds = segmentStartOffset.TotalSeconds,
SegmentDurationSeconds = frameCount * converter.SecondsPerFrameStep,
Name = "noName",
};
events.Add(ae);
// fill the intensity array
var startRow = converter.FrameFromStartTime(track.StartTimeSeconds);
var amplitudeTrack = track.GetAmplitudeOverTimeFrames();
for (int i = 0; i < amplitudeTrack.Length; i++)
{
temporalIntensityArray[startRow + i] += amplitudeTrack[i];
}
}
// MAY NOT WANT TO Do THIS FOR ONE-FRAME tracks
// combine proximal events that occupy similar frequency band
//if (combineProximalSimilarEvents)
//{
// TimeSpan startDifference = TimeSpan.FromSeconds(0.5);
// int hertzDifference = 500;
// //######################################################################## TODO TODO TODOD
// //events = AcousticEvent.CombineSimilarProximalEvents(events, startDifference, hertzDifference);
//}
return(events, temporalIntensityArray);
}
public void TestOneframeTrackAlgorithm()
{
// Set up the recognizer parameters.
var parameters = new OneframeTrackParameters()
{
MinHertz = 6000,
MaxHertz = 11000,
MinBandwidthHertz = 100,
MaxBandwidthHertz = 5000,
DecibelThreshold = 2.0,
};
//Set up the virtual recording.
int samplerate = 22050;
double signalDuration = 13.0; //seconds
// set up the config for a virtual spectrogram.
var sonoConfig = new SonogramConfig()
{
WindowSize = 512,
WindowStep = 512,
WindowOverlap = 0.0, // this must be set
WindowFunction = WindowFunctions.HANNING.ToString(),
NoiseReductionType = NoiseReductionType.Standard,
NoiseReductionParameter = 0.0,
Duration = TimeSpan.FromSeconds(signalDuration),
SampleRate = samplerate,
};
var spectrogram = this.CreateArtificialSpectrogramToTestTracksAndHarmonics(sonoConfig);
//var image1 = SpectrogramTools.GetSonogramPlusCharts(spectrogram, null, null, null);
//results.Sonogram.GetImage().Save(this.outputDirectory + "\\debug.png");
var segmentStartOffset = TimeSpan.Zero;
var plots = new List <Plot>();
var(spectralEvents, dBArray) = OneframeTrackAlgorithm.GetOneFrameTracks(
spectrogram,
parameters,
segmentStartOffset);
// draw a plot of max decibels in each frame
double decibelNormalizationMax = 5 * parameters.DecibelThreshold.Value;
var dBThreshold = parameters.DecibelThreshold.Value / decibelNormalizationMax;
var normalisedDecibelArray = DataTools.NormaliseInZeroOne(dBArray, 0, decibelNormalizationMax);
var plot1 = new Plot("decibel max", normalisedDecibelArray, dBThreshold);
plots.Add(plot1);
var allResults = new RecognizerResults()
{
NewEvents = new List <EventCommon>(),
Hits = null,
ScoreTrack = null,
Plots = new List <Plot>(),
Sonogram = null,
};
// combine the results i.e. add the events list of call events.
allResults.NewEvents.AddRange(spectralEvents);
allResults.Plots.AddRange(plots);
// effectively keeps only the *last* sonogram produced
allResults.Sonogram = spectrogram;
// DEBUG PURPOSES COMMENT NEXT LINE
this.SaveTestOutput(
outputDirectory => GenericRecognizer.SaveDebugSpectrogram(allResults, null, outputDirectory, "ClickTrack"));
Assert.AreEqual(6, allResults.NewEvents.Count);
var @event = (SpectralEvent)allResults.NewEvents[0];
Assert.AreEqual(10.0, @event.EventStartSeconds, 0.1);
Assert.AreEqual(10.1, @event.EventEndSeconds, 0.1);
Assert.AreEqual(6450, @event.LowFrequencyHertz);
Assert.AreEqual(10750, @event.HighFrequencyHertz);
@event = (SpectralEvent)allResults.NewEvents[2];
Assert.AreEqual(11.05, @event.EventStartSeconds, 0.05);
Assert.AreEqual(11.07, @event.EventEndSeconds, 0.05);
Assert.AreEqual(6450, @event.LowFrequencyHertz);
Assert.AreEqual(7310, @event.HighFrequencyHertz);
}
