Skip to content

Commit

Permalink
Update CrossCorrelation.cs
Browse files Browse the repository at this point in the history 
Issue #471 Removed the only useful method from this class and placed in HarmonicPArameters.cs.
THis CrossCorrelations.cs class is now redundant and could possibly be removed except that it contains methods previously used in recognition of crow calls and human speech. However I doubt they of use any longer.
  • Loading branch information
@towsey
towsey committed Jun 13, 2021
1 parent 6abe578 commit d873b10
Showing 1 changed file with 23 additions and 114 deletions.
137 changes: 23 additions & 114 deletions src/AudioAnalysisTools/CrossCorrelation.cs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,28 +5,34 @@
namespace AudioAnalysisTools
{
using System;
using Accord.Math;
using AudioAnalysisTools.DSP;
using TowseyLibrary;

/// <summary>
/// This class contains two methods that could eventually be deleted.
/// The methods are only called by call recognizers that have not been maintained in a long time.
/// </summary>
public class CrossCorrelation
{
// THESE KEYS COMMENTED 2021 June 13 as they appear to be unused.
//these keys are used to define a cross-correlation event in a sonogram.
public const string key_COUNT = "count";
public const string key_START_FRAME = "startFrame";
public const string key_END_FRAME = "endFrame";
public const string key_FRAME_COUNT = "frameCount";
public const string key_START_SECOND = "startSecond";
public const string key_END_SECOND = "endSecond";
public const string key_MIN_FREQBIN = "minFreqBin";
public const string key_MAX_FREQBIN = "maxFreqBin";
public const string key_MIN_FREQ = "minFreq";
public const string key_MAX_FREQ = "maxFreq";
public const string key_SCORE = "score";
public const string key_PERIODICITY = "periodicity";
//public const string key_COUNT = "count";
//public const string key_START_FRAME = "startFrame";
//public const string key_END_FRAME = "endFrame";
//public const string key_FRAME_COUNT = "frameCount";
//public const string key_START_SECOND = "startSecond";
//public const string key_END_SECOND = "endSecond";
//public const string key_MIN_FREQBIN = "minFreqBin";
//public const string key_MAX_FREQBIN = "maxFreqBin";
//public const string key_MIN_FREQ = "minFreq";
//public const string key_MAX_FREQ = "maxFreq";
//public const string key_SCORE = "score";
//public const string key_PERIODICITY = "periodicity";

/// <summary>
/// This method assume the matrix is derived from a spectrogram rotated so that the matrix rows are spectral columns of sonogram.
/// TODO THis method could eventually be deleted. It has been replaced by the other methods below.
/// Amongst other things, the term "periodicity" is used incorrectly in this method.
/// It actually refers to the "harmonic interval".
/// This method assumes the matrix is derived from a spectrogram rotated so that the matrix rows are spectral timeframes of a spectrogram.
///
/// </summary>
public static Tuple<double[], double[]> DetectBarsInTheRowsOfaMatrix(double[,] m, double threshold, int zeroBinCount)
Expand Down Expand Up @@ -71,6 +77,8 @@ public static Tuple<double[], double[]> DetectBarsInTheRowsOfaMatrix(double[,] m
} //DetectBarsInTheRowsOfaMatrix()

/// <summary>
/// TODO TODO this method could be deleted. It is called only by a method to detect crow calls.
/// THis is long since superceded.
/// A METHOD TO DETECT HARMONICS IN THE ROWS of the passed portion of a sonogram.
/// This method assume the matrix is derived from a spectrogram rotated so that the matrix rows are spectral columns of sonogram.
/// Was first developed for crow calls.
Expand Down Expand Up @@ -122,104 +130,5 @@ public static Tuple<double[], double[], double[]> DetectHarmonicsInSonogramMatri

return Tuple.Create(dBArray, intensity, periodicity);
}

/// <summary>
/// A METHOD TO DETECT a set of stacked HARMONICS/FORMANTS in the sub-band of a spectrogram.
/// Developed for GenericRecognizer of harmonics.
/// NOTE 1: This method assumes the matrix is derived from a spectrogram rotated so that the matrix rows are spectral columns of the spectrogram.
/// NOTE 2: As of March 2020, this method averages the values in five adjacent frames. This is to reduce noise.
/// But it requires that the frequency of any potential formants is not changing rapidly.
/// This may not be suitable for detecting human speech. However can reduce the frame step.
/// NOTE 3: This method assumes that the minimum number of formants in a stack = 3.
/// This means that the first 4 values in the returned array of DCT coefficients are set = 0 (see below).
/// </summary>
/// <param name="m">data matrix derived from the subband of a spectrogram.</param>
/// <param name="xThreshold">Minimum acceptable value to be considered part of a harmonic.</param>
/// <returns>three arrays: dBArray, intensity, maxIndexArray.</returns>
public static Tuple<double[], double[], int[]> DetectHarmonicsInSpectrogramData(double[,] m, double xThreshold)
{
int rowCount = m.GetLength(0);
int colCount = m.GetLength(1);
var binCount = m.GetLength(1);

//set up the cosine coefficients
double[,] cosines = MFCCStuff.Cosines(binCount, binCount);

// set up time-frame arrays to store decibels, formant intensity and max index.
var dBArray = new double[rowCount];
var intensity = new double[rowCount];
var maxIndexArray = new int[rowCount];

// for all time frames
for (int t = 2; t < rowCount - 2; t++)
{
// Smooth the frame values by taking the average of five adjacent frames
var frame1 = MatrixTools.GetRow(m, t - 2);
var frame2 = MatrixTools.GetRow(m, t - 1);
var frame3 = MatrixTools.GetRow(m, t);
var frame4 = MatrixTools.GetRow(m, t + 1);
var frame5 = MatrixTools.GetRow(m, t + 2);

// set up a frame of average db values.
var avFrame = new double[colCount];
for (int i = 0; i < colCount; i++)
{
//avFrame[i] = (frame2[i] + frame3[i] + frame4[i]) / 3;
avFrame[i] = (frame1[i] + frame2[i] + frame3[i] + frame4[i] + frame5[i]) / 5;
}

// ignore frame if its maximum decibel value is below the threshold.
double maxValue = avFrame.Max();
dBArray[t] = maxValue;
if (maxValue < xThreshold)
{
continue;
}

// do the autocross-correlation prior to doing the DCT.
double[] xr = AutoAndCrossCorrelation.AutoCrossCorr(avFrame);

// xr has twice length of frame and is symmetrical. Require only first half.
double[] normXr = new double[colCount];
for (int i = 0; i < colCount; i++)
{
// Typically normalise the xcorr values for overlap count.
// i.e. normXr[i] = xr[i] / (colCount - i);
// But for harmonics, this introduces too much noise - need to give less weight to the less overlapped values.
// Therefore just normalise by dividing values by the first, so first value = 1.
normXr[i] = xr[i] / xr[0];
//normXr[i] = xr[i];
}

//normXr = DataTools.DiffFromMean(normXr);

// fit the x-correlation array to a line to remove first order trend.
// This will help in detecting the correct maximum DCT coefficient.
var xValues = new double[normXr.Length];
for (int j = 0; j < xValues.Length; j++)
{ xValues[j] = j; }

// get the line of best fit and subtract to get deviation from the line.
Tuple<double, double> values = MathNet.Numerics.Fit.Line(xValues, normXr);
var intercept = values.Item1;
var slope = values.Item2;
for (int j = 0; j < xValues.Length; j++)
{
var lineValue = (slope * j) + intercept;
normXr[j] -= lineValue;
}

// now do DCT across the detrended auto-cross-correlation
// set the first four values in the returned DCT coefficients to 0.
// We require a minimum of three formants, that is two gaps.
int lowerDctBound = 4;
var dctCoefficients = Oscillations2012.DoDct(normXr, cosines, lowerDctBound);
int indexOfMaxValue = DataTools.GetMaxIndex(dctCoefficients);
intensity[t] = dctCoefficients[indexOfMaxValue];
maxIndexArray[t] = indexOfMaxValue;
}

return Tuple.Create(dBArray, intensity, maxIndexArray);
}
}
}

0 comments on commit d873b10

Please sign in to comment.