Class NoiseRemoval_Briggs

Inheritance

Object

NoiseRemoval_Briggs

Inherited Members

Object.Equals(Object)

Object.Equals(Object, Object)

Object.GetHashCode()

Object.GetType()

Object.MemberwiseClone()

Object.ReferenceEquals(Object, Object)

Object.ToString()

Namespace: AudioAnalysisTools.DSP

Assembly: AudioAnalysisTools.dll

Syntax

public static class NoiseRemoval_Briggs

Methods

BriggsNoiseFilterAndGetMask(Double[,], Int32, Double)

Declaration

public static double[, ] BriggsNoiseFilterAndGetMask(double[, ] matrix, int percentileThreshold, double binaryThreshold)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	percentileThreshold
Double	binaryThreshold

Returns

Type	Description
Double[,]

BriggsNoiseFilterAndGetSonograms(Double[,], Int32, Double, TimeSpan, TimeSpan, TimeSpan, Int32, Int32)

Declaration

public static Image BriggsNoiseFilterAndGetSonograms(double[, ] matrix, int percentileThreshold, double binaryThreshold, TimeSpan recordingDuration, TimeSpan xAxisInterval, TimeSpan stepDuration, int nyquist, int herzInterval)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	percentileThreshold
Double	binaryThreshold
TimeSpan	recordingDuration
TimeSpan	xAxisInterval
TimeSpan	stepDuration
Int32	nyquist
Int32	herzInterval

Returns

Type	Description
SixLabors.ImageSharp.Image

CalculateLocalVariance1(Double[], Int32)

This was written for the local contrast normalisation (LCN) of amplitude spectrograms. However the contrast is calculated wrt the local part of frequency bin or column. Plugging up ends of the returned array as done here is a hack but it does not really matter.

Declaration

public static double[] CalculateLocalVariance1(double[] data, int window)

Parameters

Type	Name	Description
Double[]	data
Int32	window

Returns

Type	Description
Double[]

CalculateLocalVariance2(Double[,], Int32, Int32)

THis method is equivalent to the above method - CalculateLocalVariance1(), except that the local variance is derived from a local matrix rather than the local frequency bin.

Declaration

public static double[] CalculateLocalVariance2(double[, ] matrix, int colId, int rowWindow)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	colId
Int32	rowWindow

Returns

Type	Description
Double[]

DrawSonogram(Double[,], TimeSpan, TimeSpan, TimeSpan, Int32, Int32, String)

Declaration

public static Image<Rgb24> DrawSonogram(double[, ] data, TimeSpan recordingDuration, TimeSpan xInterval, TimeSpan xAxisPixelDuration, int nyquist, int herzInterval, string title)

Parameters

Type	Name	Description
Double[,]	data
TimeSpan	recordingDuration
TimeSpan	xInterval
TimeSpan	xAxisPixelDuration
Int32	nyquist
Int32	herzInterval
String	title

Returns

Type	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>

NoiseReductionByDivision(Double[,], Int32)

Assumes the passed matrix is a spectrogram. i.e. rows=frames, cols=freq bins. WARNING: This method should NOT be used for short recordings (i.e LT approx 10-15 seconds long) Obtains a background noise profile from the passed percentile of lowest energy frames, Then divide cell energy by the profile value. This method was adapted from a paper by Briggs.

Declaration

public static double[, ] NoiseReductionByDivision(double[, ] matrix, int percentileThreshold)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	percentileThreshold

Returns

Type	Description
Double[,]

NoiseReductionByDivisionAndSqrRoot(Double[,], Int32)

Assumes the passed matrix is a spectrogram. i.e. rows=frames, cols=freq bins. WARNING: This method should NOT be used for short recordings (i.e LT approx 10-15 seconds long) because it obtains a background noise profile from the passed percentile of lowest energy frames.

Same method as above except take square root of the cell energy divided by the noise. Taking the square root has the effect of reducing image contrast.

Declaration

public static double[, ] NoiseReductionByDivisionAndSqrRoot(double[, ] matrix, int percentileThreshold)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	percentileThreshold

Returns

Type	Description
Double[,]

NoiseReductionByLcn(Double[,], Int32, Double)

THis method similar to the above BUT: 1: it does not do initial subtraction of lowest percentile noise. 2: it calculates local variance from local matrix andnot local frequency bin. Assumes the passed matrix is a spectrogram. i.e. rows=frames, cols=freq bins. Currently, the denominator = (contrastLevel + Math.Sqrt(localVariance[y]) A low contrastLevel = 0.1 give more grey image. A high contrastLevel = 1.0 give mostly white high contrast image. I tried various other normalisation equations as can be seen below. NOTE: Taking square-root of top line results in too much background. The algorithm is not overly sensitive to the neighbourhood size.

Declaration

public static double[, ] NoiseReductionByLcn(double[, ] matrix, int neighbourhood, double contrastLevel)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	neighbourhood	suitable vaues are odd numbers 9 - 59.
Double	contrastLevel	Suitable values are 0.1 to 1.0.

Returns

Type	Description
Double[,]

NoiseReductionByLcn(Double[,], Int32, Int32, Double)

Assumes the passed matrix is a spectrogram. i.e. rows=frames, cols=freq bins. First obtains background noise profile calculated from lowest 20% of cells for each freq bin independently. Loop over freq bins (columns) - subtract noise and divide by LCN (Local Contrast Normalisation.

The LCN denominator = (contrastLevelConstant + Sqrt(localVariance[y]) Note that sqrt of variance = std dev. A low contrastLevel = 0.1 give more grey image. A high contrastLevel = 1.0 give mostly white high contrast image.

Declaration

public static double[, ] NoiseReductionByLcn(double[, ] matrix, int lowPercent, int neighbourhood, double contrastLevel)

Parameters

Type	Name	Description
Double[,]	matrix
Int32	lowPercent
Int32	neighbourhood
Double	contrastLevel

Returns

Type	Description
Double[,]

NoiseReductionByLowestPercentileSubtraction(Double[,], Int32)

Assumes the passed matrix is a spectrogram. i.e. rows=frames, cols=freq bins, and that all values in data matrix are positive. WARNING: This method should NOT be used for short recordings (i.e LT approx 10-15 seconds long) Obtains a background noise profile from the passed percentile of lowest energy frames, Then subtracts the noise profile value from every cell. This method was adapted from a paper by Briggs.

Declaration

public static double[, ] NoiseReductionByLowestPercentileSubtraction(double[, ] matrix, int percentileThreshold)

Parameters

Type	Name	Description
Double[,]	matrix	the passed amplitude or energy spectrogram.
Int32	percentileThreshold	Must be an integer percent.

Returns

Type	Description
Double[,]	Spectrogram data matrix with noise subtracted.