public static void CropEvents(List <AcousticEvent> events, double[] activity, double minDurationInSeconds)
{
double severity = 0.2;
int length = activity.Length;
foreach (AcousticEvent ev in events)
{
int start = ev.oblong.r1;
int end = ev.oblong.r2;
double[] subArray = DataTools.Subarray(activity, start, end - start + 1);
int[] bounds = DataTools.Peaks_CropLowAmplitude(subArray, severity);
int newMinRow = start + bounds[0];
int newMaxRow = start + bounds[1];
if (newMaxRow >= length)
{
newMaxRow = length - 1;
}
Oblong o = new Oblong(newMinRow, ev.oblong.c1, newMaxRow, ev.oblong.c2);
ev.oblong = o;
ev.TimeStart = newMinRow * ev.FrameOffset;
ev.TimeEnd = newMaxRow * ev.FrameOffset;
ev.Duration = ev.TimeEnd - ev.TimeStart;
//int frameCount = (int)Math.Round(ev.Duration / ev.FrameOffset);
}
for (int i = events.Count - 1; i >= 0; i--)
{
if (events[i].Duration < minDurationInSeconds)
{
events.Remove(events[i]);
}
}
}
public static List <AcousticEvent> ReadGroundParrotTemplateAsList(BaseSonogram sonogram)
{
var timeScale = sonogram.FrameStep;
var hzScale = (int)sonogram.FBinWidth;
int rows = GroundParrotTemplate1.GetLength(0);
int cols = GroundParrotTemplate1.GetLength(1);
double timeOffset = GroundParrotTemplate1[0, 0];
var gpTemplate = new List <AcousticEvent>();
for (int r = 0; r < rows; r++)
{
int t1 = (int)Math.Round((GroundParrotTemplate1[r, 0] - timeOffset) / timeScale);
int t2 = (int)Math.Round((GroundParrotTemplate1[r, 1] - timeOffset) / timeScale);
int f2 = (int)Math.Round(GroundParrotTemplate1[r, 2] / hzScale);
int f1 = (int)Math.Round(GroundParrotTemplate1[r, 3] / hzScale);
Oblong o = new Oblong(t1, f1, t2, f2);
gpTemplate.Add(
new AcousticEvent(
TimeSpan.Zero,
o,
sonogram.NyquistFrequency,
sonogram.Configuration.FreqBinCount,
sonogram.FrameDuration,
sonogram.FrameStep,
sonogram.FrameCount));
}
return(gpTemplate);
}
} //Analysis()
public static void CropEvents(List <AcousticEvent> events, double[] intensity)
{
double severity = 0.1;
int length = intensity.Length;
foreach (AcousticEvent ev in events)
{
int start = ev.Oblong.RowTop;
int end = ev.Oblong.RowBottom;
double[] subArray = DataTools.Subarray(intensity, start, end - start + 1);
int[] bounds = DataTools.Peaks_CropLowAmplitude(subArray, severity);
int newMinRow = start + bounds[0];
int newMaxRow = start + bounds[1];
if (newMaxRow >= length)
{
newMaxRow = length - 1;
}
Oblong o = new Oblong(newMinRow, ev.Oblong.ColumnLeft, newMaxRow, ev.Oblong.ColumnRight);
ev.Oblong = o;
ev.TimeStart = newMinRow * ev.FrameOffset;
ev.TimeEnd = newMaxRow * ev.FrameOffset;
}
}
private static bool EnumMonitorsProc(IntPtr hMonitor, IntPtr hdcMonitor, ref RectStruct lprcMonitor, IntPtr dwData)
{
MonitorInfoEx mi = new MonitorInfoEx();
mi.Size = (uint)Marshal.SizeOf(mi);
bool success = GetMonitorInfo(hMonitor, ref mi);
if (success)
{
ScreenInfo si = new ScreenInfo();
si.MonitorArea = Oblong.FromRectStruct(mi.Monitor);
si.WorkArea = Oblong.FromRectStruct(mi.WorkArea);
si.DeviceName = mi.DeviceName;
si.IsPrimaryScreen = ((mi.Flags & MONITORINFOF_PRIMARY) == 1);;
DEVMODE DeviceMode = new DEVMODE();
DeviceMode.Initialize();
if (EnumDisplaySettingsEx(ToLPTStr(mi.DeviceName), -1, ref DeviceMode))
{
si.Scaling = Math.Round(((double)DeviceMode.dmPelsHeight / (mi.Monitor.bottom - mi.Monitor.top)) * 100);
}
si.NativeWorkArea = new Oblong((int)(mi.WorkArea.left * si.Scaling) / 100, (int)(mi.WorkArea.top * si.Scaling) / 100, (int)(mi.WorkArea.right * si.Scaling) / 100, (int)(mi.WorkArea.bottom * si.Scaling) / 100);
si.NativeArea = new Oblong((int)(mi.Monitor.left * si.Scaling) / 100, (int)(mi.Monitor.top * si.Scaling) / 100, (int)(mi.Monitor.right * si.Scaling) / 100, (int)(mi.Monitor.bottom * si.Scaling) / 100);
Display.Screens.Add(si);
}
return(true);
}
public void SetBounds <T>(T @event, Oblong source)
where T : SpectralEvent
{
@event.EventStartSeconds = this.GetStartTimeInSecondsOfFrame(source.RowTop);
@event.EventEndSeconds = this.GetEndTimeInSecondsOfFrame(source.RowBottom);
@event.LowFrequencyHertz = this.GetHertzFromFreqBin(source.ColumnLeft);
@event.HighFrequencyHertz = this.GetHertzFromFreqBin(source.ColumnRight);
}
public static shape getSgape(string flag)
{
shape area;
switch (flag)
{
case "1": area = new Square(); return(area);
case "2": area = new Oblong(); return(area);
case "3": area = new Triangle(); return(area);
default: Console.WriteLine("没有所选的形状。"); return(null);
}
}
public static double[] DetectEPR(List <AcousticEvent> template, BaseSonogram sonogram, double[] odScores, double odThreshold)
{
int length = sonogram.FrameCount;
double[] eprScores = new double[length];
Oblong ob1 = template[0].Oblong; // the first chirp in template
Oblong obZ = template[template.Count - 1].Oblong; // the last chirp in template
int templateLength = obZ.RowBottom;
for (int frame = 0; frame < length - templateLength; frame++)
{
if (odScores[frame] < odThreshold)
{
continue;
}
// get best freq band and max score for the first rectangle.
double maxScore = -double.MaxValue;
int freqBinOffset = 0;
for (int bin = -5; bin < 15; bin++)
{
Oblong ob = new Oblong(ob1.RowTop + frame, ob1.ColumnLeft + bin, ob1.RowBottom + frame, ob1.ColumnRight + bin);
double score = GetLocationScore(sonogram, ob);
if (score > maxScore)
{
maxScore = score;
freqBinOffset = bin;
}
}
//if location score exceeds threshold of 6 dB then get remaining scores.
if (maxScore < 6.0)
{
continue;
}
foreach (AcousticEvent ae in template)
{
Oblong ob = new Oblong(ae.Oblong.RowTop + frame, ae.Oblong.ColumnLeft + freqBinOffset, ae.Oblong.RowBottom + frame, ae.Oblong.ColumnRight + freqBinOffset);
double score = GetLocationScore(sonogram, ob);
eprScores[frame] += score;
}
eprScores[frame] /= template.Count;
}
return(eprScores);
}
public override ShapeBase Clone()
{
var shape = new Oblong
{
Location = Location,
Size = Size,
DrawMethod = DrawMethod,
OutlineColor = OutlineColor,
OutlineWidth = OutlineWidth,
FillColor = FillColor,
IsClosedFigure = IsClosedFigure,
Vertices = (VertexCollection)Vertices.Clone()
};
return(shape);
}
/// <summary>
/// reutrns the difference between the maximum dB value in a retangular location and the average of the boundary dB values.
/// </summary>
/// <param name="sonogram"></param>
/// <param name="ob"></param>
/// <returns></returns>
public static double GetLocationScore(BaseSonogram sonogram, Oblong ob)
{
double max = -double.MaxValue;
for (int r = ob.RowTop; r < ob.RowBottom; r++)
{
for (int c = ob.ColumnLeft; c < ob.ColumnRight; c++)
{
if (sonogram.Data[r, c] > max)
{
max = sonogram.Data[r, c];
}
}
}
//calculate average boundary value
int boundaryLength = 2 * (ob.RowBottom - ob.RowTop + 1 + ob.ColumnRight - ob.ColumnLeft + 1);
double boundaryValue = 0.0;
for (int r = ob.RowTop; r < ob.RowBottom; r++)
{
boundaryValue += sonogram.Data[r, ob.ColumnLeft] + sonogram.Data[r, ob.ColumnRight];
}
for (int c = ob.ColumnLeft; c < ob.ColumnRight; c++)
{
boundaryValue += sonogram.Data[ob.RowTop, c] + sonogram.Data[ob.RowBottom, c];
}
boundaryValue /= boundaryLength;
double score = max - boundaryValue;
if (score < 0.0)
{
score = 0.0;
}
return(score);
}
/// <summary>
/// Does the Analysis
/// Returns a DataTable
/// </summary>
/// <param name="fiSegmentOfSourceFile"></param>
/// <param name="configDict"></param>
/// <param name="diOutputDir"></param>
/// <param name="opFileName"></param>
/// <param name="segmentStartOffset"></param>
/// <param name="config"></param>
/// <param name="segmentAudioFile"></param>
public static Tuple <BaseSonogram, double[, ], double[], List <AcousticEvent>, TimeSpan> Analysis(FileInfo fiSegmentOfSourceFile, Dictionary <string, string> configDict, DirectoryInfo diOutputDir, string opFileName, TimeSpan segmentStartOffset)
{
//set default values
int bandWidth = 500; //detect bars in bands of this width.
int frameSize = 1024;
double windowOverlap = 0.0;
double intensityThreshold = double.Parse(configDict[key_INTENSITY_THRESHOLD]);
//intensityThreshold = 0.01;
AudioRecording recording = AudioRecording.GetAudioRecording(fiSegmentOfSourceFile, RESAMPLE_RATE, diOutputDir.FullName, opFileName);
if (recording == null)
{
LoggedConsole.WriteLine("############ WARNING: Recording could not be obtained - likely file does not exist.");
return(null);
}
int sr = recording.SampleRate;
double binWidth = recording.SampleRate / (double)frameSize;
double frameDuration = frameSize / (double)sr;
double frameOffset = frameDuration * (1 - windowOverlap); //seconds between start of each frame
double framesPerSecond = 1 / frameOffset;
TimeSpan tsRecordingtDuration = recording.Duration;
int colStep = (int)Math.Round(bandWidth / binWidth);
//i: GET SONOGRAM AS MATRIX
double epsilon = Math.Pow(0.5, recording.BitsPerSample - 1);
var results2 = DSP_Frames.ExtractEnvelopeAndAmplSpectrogram(recording.WavReader.Samples, sr, epsilon, frameSize, windowOverlap);
double[] avAbsolute = results2.Average; //average absolute value over the minute recording
//double[] envelope = results2.Item2;
double[,] spectrogram = results2.AmplitudeSpectrogram; //amplitude spectrogram. Note that column zero is the DC or average energy value and can be ignored.
double windowPower = results2.WindowPower;
//############################ NEXT LINE FOR DEBUGGING ONLY
//spectrogram = GetTestSpectrogram(spectrogram.GetLength(0), spectrogram.GetLength(1), 0.01, 0.03);
var output = DetectGratingEvents(spectrogram, colStep, intensityThreshold);
var amplitudeArray = output.Item2; //for debug purposes only
//convert List of Dictionary events to List of ACousticevents.
//also set up the hits matrix.
int rowCount = spectrogram.GetLength(0);
int colCount = spectrogram.GetLength(1);
var hitsMatrix = new double[rowCount, colCount];
var acousticEvents = new List <AcousticEvent>();
double minFrameCount = 8; //this assumes that the minimum grid is 2 * 4 = 8 long
foreach (Dictionary <string, double> item in output.Item1)
{
int minRow = (int)item[key_START_FRAME];
int maxRow = (int)item[key_END_FRAME];
int frameCount = maxRow - minRow + 1;
if (frameCount < minFrameCount)
{
continue; //only want events that are over a minimum length
}
int minCol = (int)item[key_MIN_FREQBIN];
int maxCol = (int)item[key_MAX_FREQBIN];
double periodicity = item[key_PERIODICITY];
double[] subarray = DataTools.Subarray(avAbsolute, minRow, maxRow - minRow + 1);
double severity = 0.1;
int[] bounds = DataTools.Peaks_CropToFirstAndLast(subarray, severity);
minRow = minRow + bounds[0];
maxRow = minRow + bounds[1];
if (maxRow >= rowCount)
{
maxRow = rowCount - 1;
}
Oblong o = new Oblong(minRow, minCol, maxRow, maxCol);
var ae = new AcousticEvent(segmentStartOffset, o, results2.NyquistFreq, frameSize, frameDuration, frameOffset, frameCount);
ae.Name = string.Format("p={0:f0}", periodicity);
ae.Score = item[key_SCORE];
ae.ScoreNormalised = item[key_SCORE] / 0.5;
acousticEvents.Add(ae);
//display event on the hits matrix
for (int r = minRow; r < maxRow; r++)
{
for (int c = minCol; c < maxCol; c++)
{
hitsMatrix[r, c] = periodicity;
}
}
} //foreach
//set up the songogram to return. Use the existing amplitude sonogram
int bitsPerSample = recording.WavReader.BitsPerSample;
//NoiseReductionType nrt = SNR.Key2NoiseReductionType("NONE");
NoiseReductionType nrt = SNR.KeyToNoiseReductionType("STANDARD");
var sonogram = (BaseSonogram)SpectrogramStandard.GetSpectralSonogram(recording.BaseName, frameSize, windowOverlap, bitsPerSample, windowPower, sr, tsRecordingtDuration, nrt, spectrogram);
sonogram.DecibelsNormalised = new double[sonogram.FrameCount];
for (int i = 0; i < sonogram.FrameCount; i++) //foreach frame or time step
{
sonogram.DecibelsNormalised[i] = 2 * Math.Log10(avAbsolute[i]);
}
sonogram.DecibelsNormalised = DataTools.normalise(sonogram.DecibelsNormalised);
return(Tuple.Create(sonogram, hitsMatrix, amplitudeArray, acousticEvents, tsRecordingtDuration));
} //Analysis()
/// <summary>
/// EXPANATION: A vertical track is a near click or rapidly frequency-modulated tone. A good example is the whip component of the whip-bird call.
/// They would typically be only a few time-frames duration.
/// THis method averages dB log values incorrectly but it is faster than doing many log conversions and is accurate enough for the purpose.
/// </summary>
public static (List <AcousticEvent> Events, double[] CombinedIntensity) GetVerticalTracks(
SpectrogramStandard sonogram,
int minHz,
int maxHz,
int nyquist,
double decibelThreshold,
int minBandwidthHertz,
int maxBandwidthHertz,
bool combineProximalSimilarEvents,
TimeSpan segmentStartOffset)
{
var sonogramData = sonogram.Data;
int frameCount = sonogramData.GetLength(0);
int binCount = sonogramData.GetLength(1);
var frameDuration = sonogram.FrameDuration;
var frameStep = sonogram.FrameStep;
var frameOverStep = frameDuration - frameStep;
double binWidth = nyquist / (double)binCount;
int minBin = (int)Math.Round(minHz / binWidth);
int maxBin = (int)Math.Round(maxHz / binWidth);
// list of accumulated acoustic events
var events = new List <AcousticEvent>();
var temporalIntensityArray = new double[frameCount];
//Find all frame peaks and place in peaks matrix
var peaks = new double[frameCount, binCount];
for (int row = 1; row < frameCount - 1; row++)
{
for (int col = minBin; col < maxBin; col++)
{
if (sonogramData[row, col] < decibelThreshold)
{
continue;
}
// if given matrix element is greater than in frame either side
bool isPeak = (sonogramData[row, col] > sonogramData[row - 1, col]) && (sonogramData[row, col] > sonogramData[row + 1, col]);
if (isPeak)
{
peaks[row, col] = sonogramData[row, col];
}
}
}
// Look for track starts and initialise them as events.
// Cannot include edge rows & columns because of edge effects.
// Each row is a time frame which is a spectrum. Each column is a frequency bin
var combinedIntensityArray = new double[frameCount];
for (int col = minBin; col < maxBin; col++)
{
for (int row = 2; row < frameCount - 2; row++)
{
// Visit each frame peak in order. Each may be start of possible track
if (peaks[row, col] < decibelThreshold)
{
continue;
}
//have the beginning of a potential track
var track = GetVerticalTrack(peaks, row, col, maxBin, decibelThreshold);
// calculate first and last of the frame IDs in the original spectrogram
int trackStartFrame = track.GetStartFrame();
int trackEndFrame = track.GetEndFrame();
// next two for debug purposes
//int trackMinBin = track.GetBottomFreqBin();
//int trackTopBin = track.GetTopFreqBin();
//If track has lies within the correct bandWidth range, then create an event
int trackBandWidth = track.GetTrackBandWidthHertz(binWidth);
if (trackBandWidth >= minBandwidthHertz && trackBandWidth <= maxBandwidthHertz)
{
// get the oblong and init an event
double trackDuration = ((trackEndFrame - trackStartFrame) * frameStep) + frameOverStep;
var oblong = new Oblong(trackStartFrame, col, trackEndFrame, track.GetTopFreqBin());
var ae = new AcousticEvent(segmentStartOffset, oblong, nyquist, binCount, frameDuration, frameStep, frameCount)
{
// get the track as matrix
TheTrack = track,
};
events.Add(ae);
// fill the intensity array
var amplitudeArray = track.GetAmplitudeOverTimeFrames();
for (int i = 0; i < amplitudeArray.Length; i++)
{
combinedIntensityArray[row + i] += amplitudeArray[i];
}
}
} // rows/frames
} // end cols/bins
// combine proximal events that occupy similar frequency band
if (combineProximalSimilarEvents)
{
TimeSpan startDifference = TimeSpan.FromSeconds(0.5);
int hertzDifference = 500;
events = AcousticEvent.CombineSimilarProximalEvents(events, startDifference, hertzDifference);
}
return(events, temporalIntensityArray);
}
/// <summary>
/// This method returns spectral peak tracks enclosed in acoustic events.
/// It averages dB log values incorrectly but it is faster than doing many log conversions.
/// </summary>
public static (List <AcousticEvent> Events, double[] CombinedIntensity) GetSpectralPeakTracks(
SpectrogramStandard sonogram,
int minHz,
int maxHz,
int nyquist,
double decibelThreshold,
double minDuration,
double maxDuration,
bool combinePossibleHarmonics,
TimeSpan segmentStartOffset)
{
var sonogramData = sonogram.Data;
int frameCount = sonogramData.GetLength(0);
int binCount = sonogramData.GetLength(1);
double binWidth = nyquist / (double)binCount;
int minBin = (int)Math.Round(minHz / binWidth);
int maxBin = (int)Math.Round(maxHz / binWidth);
int bandwidthBinCount = maxBin - minBin + 1;
var frameDuration = sonogram.FrameDuration;
var frameStep = sonogram.FrameStep;
var frameOverStep = frameDuration - frameStep;
// list of accumulated acoustic events
var events = new List <AcousticEvent>();
//Find all spectral peaks and place in peaks matrix
var peaks = new double[frameCount, bandwidthBinCount];
for (int row = 0; row < frameCount; row++)
{
for (int col = minBin - 1; col < maxBin - 1; col++)
{
if (sonogramData[row, col] < decibelThreshold)
{
continue;
}
// if given matrix element is greater than in freq bin either side
bool isPeak = (sonogramData[row, col] > sonogramData[row, col - 1]) && (sonogramData[row, col] > sonogramData[row, col + 1]);
if (isPeak)
{
peaks[row, col] = sonogramData[row, col];
}
}
}
// Look for track starts and initialise them as events.
// Cannot include edge rows & columns because of edge effects.
// Each row is a time frame which is a spectrum. Each column is a frequency bin
var combinedIntensityArray = new double[frameCount];
for (int row = 0; row < frameCount; row++)
{
for (int col = 3; col < bandwidthBinCount - 3; col++)
{
// Visit each spectral peak in order. Each may be start of possible track
if (peaks[row, col] < decibelThreshold)
{
continue;
}
//have the beginning of a potential track
var track = GetTrack(peaks, row, col, decibelThreshold);
int trackStartFrame = track.GetStartFrame();
int trackEndFrame = track.GetEndFrame();
double trackDuration = ((trackEndFrame - trackStartFrame) * frameStep) + frameOverStep;
// calculate max and min bin IDs in the original spectrogram
int trackBottomBin = track.GetBottomFreqBin();
int trackTopBin = track.GetTopFreqBin();
//If track has length within duration bounds, then create an event
if (trackDuration >= minDuration && trackDuration <= maxDuration)
{
var oblong = new Oblong(track.GetStartFrame(), trackBottomBin, track.GetEndFrame(), trackTopBin);
var ae = new AcousticEvent(segmentStartOffset, oblong, nyquist, binCount, frameDuration, frameStep, frameCount)
{
// get the track as matrix
TheTrack = track,
};
events.Add(ae);
// fill the intensity array
var amplitudeTrack = track.GetAmplitudeOverTimeFrames();
for (int i = 0; i < amplitudeTrack.Length; i++)
{
combinedIntensityArray[row + i] += amplitudeTrack[i];
}
}
}
}
// Combine coincident events that are stacked one above other.
// This will help in some cases to combine related events.
var startDifference = TimeSpan.FromSeconds(0.2);
var hertzGap = 200;
if (combinePossibleHarmonics)
{
events = AcousticEvent.CombinePotentialStackedTracks(events, startDifference, hertzGap);
}
return(events, combinedIntensityArray);
}
