Skip to content

Commit

Permalink
Mean, Median and Modal noise removal
Browse files Browse the repository at this point in the history 
Write code to compare mean, meidan and modal noise removal from standard BAC environmental recording.
No difference can be seen in the output noise reduced spectrograms.
dB noise value in bin 100 of Median noise = -91.46
dB noise value in bin 100 of Mean    noise = -90.74
dB noise value in bin 100 of Modal   noise = -91.61
  • Loading branch information
@towsey
towsey committed Apr 3, 2018
1 parent b4c1e18 commit 7112ad6
Show file tree
Hide file tree
Showing 4 changed files with 196 additions and 125 deletions.
212 changes: 115 additions & 97 deletions src/AnalysisPrograms/Sandpit.cs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -85,14 +85,14 @@ public override Task<int> Execute(CommandLineApplication app)
//TEST_FilterMovingAverage();
//TestImageProcessing();
//TestMatrix3dClass();
TestsOfFrequencyScales();
//TestsOfFrequencyScales();
//TestReadingFileOfSummaryIndices();
//TestStructureTensor();
//TestWavelets();
//TestFft2D();
//TestTernaryPlots();
//TestDirectorySearchAndFileSearch();
//TestNoiseReduction();
TestNoiseReduction();

Console.WriteLine("# Finished Sandpit Task! Press any key to exit.");
return this.Ok();
Expand All @@ -116,7 +116,9 @@ public static void DrawClusterSequence1(int clusterId)
var g = Graphics.FromImage(image);
g.Clear(Color.White);

DirectoryInfo dir = new DirectoryInfo(@"H:\Documents\SensorNetworks\Students\AniekRoelofs\Results 22Feb2018\Complete\SummaryIndices");
DirectoryInfo dir =
new DirectoryInfo(
@"H:\Documents\SensorNetworks\Students\AniekRoelofs\Results 22Feb2018\Complete\SummaryIndices");
string fileName = "Cluster60Sequence_Gympie2015June22_SuI.csv";
string path = Path.Combine(dir.FullName, fileName);

Expand All @@ -136,8 +138,7 @@ public static void DrawClusterSequence1(int clusterId)
int minuteId = counter % width;
image.SetPixel(minuteId, day, Color.Red);
}
else
if (clusterId == int.Parse(words[0]))
else if (clusterId == int.Parse(words[0]))
{
int day = counter / width;
int minuteId = int.Parse(words[1]);
Expand All @@ -164,12 +165,17 @@ public static void DrawClusterSequence2(int requiredCusterId)
int height = 400;
int width = 1440;

Color[] colorchart = { Color.Red, Color.Orange, Color.Green, Color.Aqua, Color.Blue, Color.BlueViolet, Color.Black, Color.DarkMagenta, Color.LawnGreen, Color.DarkRed};
Color[] colorchart =
{
Color.Red, Color.Orange, Color.Green, Color.Aqua, Color.Blue, Color.BlueViolet, Color.Black,
Color.DarkMagenta, Color.LawnGreen, Color.DarkRed
};
var image = new Bitmap(width, height, PixelFormat.Format24bppRgb);
var g = Graphics.FromImage(image);
g.Clear(Color.White);

DirectoryInfo dir = new DirectoryInfo(@"H:\Documents\SensorNetworks\Students\AniekRoelofs\Results 14March2018");
DirectoryInfo dir =
new DirectoryInfo(@"H:\Documents\SensorNetworks\Students\AniekRoelofs\Results 14March2018");
string fileName = "Cluster60_Hidstate10_Gympie2015June22_SuI_bestof4.csv";
string path = Path.Combine(dir.FullName, fileName);
int clusterSize = 0;
Expand All @@ -191,8 +197,7 @@ public static void DrawClusterSequence2(int requiredCusterId)
{
image.SetPixel(minuteOfday, day, Color.Gray);
}
else
if (requiredCusterId == int.Parse(words[1]))
else if (requiredCusterId == int.Parse(words[1]))
{
clusterSize++;
image.SetPixel(minuteOfday, day, colorchart[hidStateId - 1]);
Expand All @@ -204,7 +209,8 @@ public static void DrawClusterSequence2(int requiredCusterId)
Console.WriteLine("Counter=" + minCounter);
} //using

g.DrawString("Cluster" + requiredCusterId + " Size=" + clusterSize, new Font("Tahoma", 20), Brushes.Black, new PointF(10, 10));
g.DrawString("Cluster" + requiredCusterId + " Size=" + clusterSize, new Font("Tahoma", 20), Brushes.Black,
new PointF(10, 10));

string imageName = $"cluster{requiredCusterId}.png";
string savePath = Path.Combine(dir.FullName + "\\ClusterImages", imageName);
Expand Down Expand Up @@ -1008,7 +1014,7 @@ public static void ExtractSpectralFeatures()
string dir =
@"H:\Documents\SensorNetworks\MyPapers\2017_DavidWatson\CaseStudy1 Liz\MachineLearningExercise";
string fileName = "LizZnidersic_TasmanIsTractor_20151111__Towsey.Acoustic";
string[] indexNames = { "ACI", "ENT", "POW", "SPT", "RHZ" };
string[] indexNames = { "ACI", "ENT", "POW", "SPT", "RHZ"};
var framecount =
1440; // could read this from first matrix but easier to declare it. Need it for reading in tagged data.
int startOffsetMinute = 47; // 24 hours of recording starts at 12:47am. Need this as an offset.
Expand Down Expand Up @@ -1209,7 +1215,7 @@ public static void TestsOfFrequencyScales()
{
// FrequencyScale.TESTMETHOD_LinearFrequencyScaleDefault();
// FrequencyScale.TESTMETHOD_LinearFrequencyScale();
FrequencyScale.TESTMETHOD_MelFrequencyScale();
// FrequencyScale.TESTMETHOD_MelFrequencyScale();
// FrequencyScale.TESTMETHOD_OctaveFrequencyScale1();
// FrequencyScale.TESTMETHOD_OctaveFrequencyScale2();

Expand Down Expand Up @@ -1237,7 +1243,7 @@ public static void TestAnalyseLongRecordingUsingArtificialSignal()
{
int sampleRate = 22050;
double duration = 420; // signal duration in seconds = 7 minutes
int[] harmonics = { 500, 1000, 2000, 4000, 8000 };
int[] harmonics = { 500, 1000, 2000, 4000, 8000};
var recording = DspFilters.GenerateTestRecording(sampleRate, duration, harmonics, WaveType.Consine);
var outputDirectory = @"C:\SensorNetworks\SoftwareTests\TestLongDurationRecordings";
var recordingPath = Path.Combine(outputDirectory, "TemporaryRecording.wav");
Expand Down Expand Up @@ -1346,7 +1352,7 @@ public static void TestArbimonSegmentationAlgorithm()
SmallAreaThreshold = 100,
};

double[] thresholdLevels = { 30.0, 25.0, 20.0, 15.0, 10.0, 5.0 };
double[] thresholdLevels = { 30.0, 25.0, 20.0, 15.0, 10.0, 5.0};
var imageList = new List<Image>();

foreach (double th in thresholdLevels)
Expand Down Expand Up @@ -1534,7 +1540,7 @@ public static void TestFft2D()
public static void TestTernaryPlots()
{
//string[] keys = { "BGN", "POW", "EVN"};
string[] keys = { "ACI", "ENT", "EVN" };
string[] keys = { "ACI", "ENT", "EVN"};

FileInfo[] indexFiles =
{
Expand Down Expand Up @@ -1697,7 +1703,7 @@ public static void ConcatenateMarineImages()

//double[] bounds = { 0.0, 3.0, 6.0 };
//double[] bounds = { 0.0, 2.0, 4.0, 8.0 };
double[] bounds = { 0.0, 2.0, 4.0, 6.0, 8.0, 10.0 }; // noSkew
double[] bounds = { 0.0, 2.0, 4.0, 6.0, 8.0, 10.0}; // noSkew

//double[] bounds = { 0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 10.0 };
//double[] bounds = { 0.0, 1.0, 2.0, 4.0, 6.0, 10.0 }; // skew left
Expand Down Expand Up @@ -1740,7 +1746,7 @@ public static void ConcatenateMarineImages()
for (int i = 0; i < fileCount; i++)
{
//ID0001_Species01.EVN.csv
char[] delimiters = { '.', 's' };
char[] delimiters = { '.', 's'};
string fileName = filePaths[i].Name;
string[] parts = fileName.Split(delimiters);
int speciesId = int.Parse(parts[1]);
Expand Down Expand Up @@ -1886,7 +1892,8 @@ public static void ResourcesForRheobatrachusSilusRecogniser()
//string recordingPath = @"C:\SensorNetworks\WavFiles\Rain\DM420036_min602.wav"; //NEGATIVE rain
//string recordingPath = @"C:\SensorNetworks\WavFiles\Noise\BAC3_20070924-153657_noise.wav"; //NEGATIVE noise
string recordingPath = @"C:\SensorNetworks\WavFiles\Frogs\Compilation6_Mono.mp3"; //FROG COMPILATION
string configPath = @"C:\SensorNetworks\Software\AudioAnalysis\AnalysisConfigFiles\Towsey.RheobatrachusSilus.cfg";
string configPath =
@"C:\SensorNetworks\Software\AudioAnalysis\AnalysisConfigFiles\Towsey.RheobatrachusSilus.cfg";
string outputDir = @"C:\SensorNetworks\Output\Frogs\";
}

Expand All @@ -1908,94 +1915,105 @@ public static void ResourcesForEventPatternRecognition()
}

/// <summary>
/// Experiments with noise reduction of spectrograms
/// THIS METHOD IS JUST A CONTAINER FOR TESTING SNIPPETS OF CODE TO DO WITH NOISE REMOVAL FROM SPECTROGRAMS
/// the following libraries are required to run these tests. They are in the SANDPIT class
/// using System.Drawing;
/// using System.Drawing.Imaging;
/// using System.IO;
/// using AudioAnalysisTools;
/// THE FOLLOWING CODE tests the use of the Noise Reduction types listed in the enum SNR.NoiseReductionType
/// The enum types include NONE, STANDARD, MODAL, Binary, etc.
/// </summary>
public static void TestNoiseReduction()
{
//#######################################################################################################################################
// experiments with noise reduction of spectrograms
//THE FOLLOWING CODE tests the use of the Noise Reduction types listed in the enum SNR.NoiseReductionType
//The enum types include NONE, STANDARD, MODAL, Binary, etc.
//COPY THE FOLLOWING CODE INTO THE CLASS Sandpit.cs to do the testing.
if (true)
//string wavFilePath = @"C:\SensorNetworks\WavFiles\LewinsRail\BAC2_20071008-085040.wav";
string wavFilePath = @"C:\SensorNetworks\WavFiles\TestRecordings\BAC\BAC2_20071008-085040.wav";
string outputDir = @"C:\SensorNetworks\Output\Test\Test2018";
//string profileName = "Median";
//string profileName = "Mean";
string profileName = "Modal";
string imageFname = "test_" + profileName + "NoiseRemoval.png";

string imagePath = Path.Combine(outputDir, imageFname);
string imageViewer = @"C:\Windows\system32\mspaint.exe";

var recording = new AudioRecording(wavFilePath);

int nyquist = recording.Nyquist;
int frameSize = 1024;
int finalBinCount = 512;
int hertzInterval = 1000;
FreqScaleType scaleType = FreqScaleType.Linear;

//FreqScaleType scaleType = FreqScaleType.Mel;
var freqScale = new FrequencyScale(scaleType, nyquist, frameSize, finalBinCount, hertzInterval);

var fst = freqScale.ScaleType;
var config = new SonogramConfig
{
//string wavFilePath = @"C:\SensorNetworks\WavFiles\LewinsRail\BAC2_20071008-085040.wav";
string wavFilePath = @"C:\SensorNetworks\WavFiles\SunshineCoast\DM420036_min407.wav";
string outputDir = @"C:\SensorNetworks\Output\Test";
string imageFname = "test3.png";

//string imagePath = Path.Combine(outputDir, imageFname);
//string imageViewer = @"C:\Windows\system32\mspaint.exe";

//var recording = new AudioRecording(wavFilePath);
//var config = new SonogramConfig { NoiseReductionType = NoiseReductionType.STANDARD, WindowOverlap = 0.0 };
//config.NoiseReductionParameter = 0.0; // backgroundNeighbourhood noise reduction in dB
//var spectrogram = new SpectralSonogram(config, recording.GetWavReader());
//Plot scores = null;
//double eventThreshold = 0.5; // dummy variable - not used
//Image image = DrawSonogram(spectrogram, scores, null, eventThreshold);
//image.Save(imagePath, ImageFormat.Png);
//FileInfo fiImage = new FileInfo(imagePath);
//if (fiImage.Exists) // Display the image using MsPaint.exe
//{
// TowseyLib.ProcessRunner process = new TowseyLib.ProcessRunner(imageViewer);
// process.Run(imagePath, outputDir);
//}
} // if(true)
NoiseReductionType = NoiseReductionType.Modal,
//NoiseReductionType = NoiseReductionType.Mean,
//NoiseReductionType = NoiseReductionType.Median,
WindowOverlap = 0.0,
WindowSize = frameSize,
};
config.NoiseReductionParameter = 0.0; // backgroundNeighbourhood noise reduction in dB

//#######################################################################################################################################
//THE FOLLOWING CODE tests the effect of changing the order of 1) CONVERT TO dB 2) NOISE REMOVAL
// versus 1) NOISE REMOVAL 2) CONVERT TO dB.
//THe results are very different. The former is GOOD. The latter is A MESS.
if (true)
// DRAW SPECTROGRAM
var spectrogram = new SpectrogramStandard(config, recording.WavReader);
var image = spectrogram.GetImageFullyAnnotated(spectrogram.GetImage(), "SPECTROGRAM: " + fst + " " + profileName, freqScale.GridLineLocations);
image.Save(imagePath, ImageFormat.Png);
FileInfo fiImage = new FileInfo(imagePath);

// Display the image using MsPaint.exe if image file exists
if (fiImage.Exists)
{
//string wavFilePath = @"C:\SensorNetworks\WavFiles\LewinsRail\BAC2_20071008-085040.wav";
string wavFilePath = @"C:\SensorNetworks\WavFiles\SunshineCoast\DM420036_min407.wav";
string outputDir = @"C:\SensorNetworks\Output\Test";
string imageFname = "test3.png";

//string imagePath = Path.Combine(outputDir, imageFname);
//string imageViewer = @"C:\Windows\system32\mspaint.exe";

//var recording = new AudioRecording(wavFilePath);
//int frameSize = 512;
//double windowOverlap = 0.0;
//// i: EXTRACT ENVELOPE and FFTs
//var results2 = DSP_Frames.ExtractEnvelopeAndFFTs(recording.GetWavReader().Samples, recording.SampleRate, frameSize, windowOverlap);

//// get amplitude spectrogram and remove the DC column ie column zero.
//double[,] spectrogramData = results2.Spectrogram;
//spectrogramData = MatrixTools.Submatrix(spectrogramData, 0, 1, spectrogramData.GetLength(0) - 1, spectrogramData.GetLength(1) - 1);
//double epsilon = Math.Pow(0.5, 16 - 1);
//double windowPower = frameSize * 0.66; //power of a rectangular window =frameSize. Hanning is less

//// convert spectrum to decibels BEFORE noise removal
////spectrogramData = Speech.DecibelSpectra(spectrogramData, windowPower, recording.SampleRate, epsilon);

//// vi: remove background noise from the spectrogram
//double SD_COUNT = 0.1;
//double SpectralBgThreshold = 0.003; // SPECTRAL AMPLITUDE THRESHOLD for smoothing background
//SNR.NoiseProfile profile = SNR.CalculateNoiseProfile(spectrogramData, SD_COUNT); //calculate noise profile - assumes a dB spectrogram.
//double[] noiseValues = DataTools.filterMovingAverage(profile.noiseThreshold, 7); // smooth the noise profile
//spectrogramData = SNR.NoiseReduce_Standard(spectrogramData, noiseValues, SpectralBgThreshold);

//// convert spectrum to decibels AFTER noise removal
////spectrogramData = Speech.DecibelSpectra(spectrogramData, windowPower, recording.SampleRate, epsilon);

//spectrogramData = MatrixTools.MatrixRotate90Anticlockwise(spectrogramData);
//ImageTools.DrawMatrix(spectrogramData, imagePath);
//FileInfo fiImage = new FileInfo(imagePath);
//if (fiImage.Exists) // Display the image using MsPaint.exe
//{
// TowseyLib.ProcessRunner process = new TowseyLib.ProcessRunner(imageViewer);
// process.Run(imagePath, outputDir);
//}
TowseyLibrary.ProcessRunner process = new TowseyLibrary.ProcessRunner(imageViewer);
process.Run(imagePath, outputDir);
}

// #################################################################################################
// THE FOLLOWING CODE tests the effect of changing the order of 1) CONVERT TO dB 2) NOISE REMOVAL
// versus 1) NOISE REMOVAL 2) CONVERT TO dB.
// THe results are very different. The former is GOOD. The latter is A MESS.
//string wavFilePath = @"C:\SensorNetworks\WavFiles\LewinsRail\BAC2_20071008-085040.wav";
//string wavFilePath = @"C:\SensorNetworks\WavFiles\SunshineCoast\DM420036_min407.wav";
//string outputDir = @"C:\SensorNetworks\Output\Test";
//string imageFname = "test3.png";

//string imagePath = Path.Combine(outputDir, imageFname);
//string imageViewer = @"C:\Windows\system32\mspaint.exe";

//var recording = new AudioRecording(wavFilePath);
//int frameSize = 512;
//double windowOverlap = 0.0;
//// i: EXTRACT ENVELOPE and FFTs
//var results2 = DSP_Frames.ExtractEnvelopeAndFFTs(recording.GetWavReader().Samples, recording.SampleRate, frameSize, windowOverlap);

//// get amplitude spectrogram and remove the DC column ie column zero.
//double[,] spectrogramData = results2.Spectrogram;
//spectrogramData = MatrixTools.Submatrix(spectrogramData, 0, 1, spectrogramData.GetLength(0) - 1, spectrogramData.GetLength(1) - 1);
//double epsilon = Math.Pow(0.5, 16 - 1);
//double windowPower = frameSize * 0.66; //power of a rectangular window =frameSize. Hanning is less

//// convert spectrum to decibels BEFORE noise removal
////spectrogramData = Speech.DecibelSpectra(spectrogramData, windowPower, recording.SampleRate, epsilon);

//// vi: remove background noise from the spectrogram
//double SD_COUNT = 0.1;
//double SpectralBgThreshold = 0.003; // SPECTRAL AMPLITUDE THRESHOLD for smoothing background
//SNR.NoiseProfile profile = SNR.CalculateNoiseProfile(spectrogramData, SD_COUNT); //calculate noise profile - assumes a dB spectrogram.
//double[] noiseValues = DataTools.filterMovingAverage(profile.noiseThreshold, 7); // smooth the noise profile
//spectrogramData = SNR.NoiseReduce_Standard(spectrogramData, noiseValues, SpectralBgThreshold);

//// convert spectrum to decibels AFTER noise removal
////spectrogramData = Speech.DecibelSpectra(spectrogramData, windowPower, recording.SampleRate, epsilon);

//spectrogramData = MatrixTools.MatrixRotate90Anticlockwise(spectrogramData);
//ImageTools.DrawMatrix(spectrogramData, imagePath);
//FileInfo fiImage = new FileInfo(imagePath);
//if (fiImage.Exists) // Display the image using MsPaint.exe
//{
// TowseyLib.ProcessRunner process = new TowseyLib.ProcessRunner(imageViewer);
// process.Run(imagePath, outputDir);
//}
}

public static void TestDct(string[] args)
Expand Down
Loading

0 comments on commit 7112ad6

Please sign in to comment.