Class SpectrogramTools

Inheritance

Object

SpectrogramTools

Inherited Members

Object.Equals(Object)

Object.Equals(Object, Object)

Object.GetHashCode()

Object.GetType()

Object.MemberwiseClone()

Object.ReferenceEquals(Object, Object)

Object.ToString()

Namespace: AudioAnalysisTools.StandardSpectrograms

Assembly: AudioAnalysisTools.dll

Syntax

public static class SpectrogramTools

Methods

AverageAnArrayOfDecibelValues(Double[])

Here is some test data for this method: array = new[] { 96.0, 100.0, 90.0, 97.0 }; The return value should = 96.988 dB First need to calculate the original value i.e. exponential or antilog. See also DataTools.AntiLogBase10(double value).

Declaration

public static double AverageAnArrayOfDecibelValues(double[] array)

Parameters

Type	Name	Description
Double[]	array	an array of decibel values.

Returns

Type	Description
Double	a decibel value.

CalculateAvgDecibelSpectrumFromDecibelSpectrogram(Double[,])

Use this method to average a decibel spectrogram.

Declaration

public static double[] CalculateAvgDecibelSpectrumFromDecibelSpectrogram(double[, ] spectrogram)

Parameters

Type	Name	Description
Double[,]	spectrogram

Returns

Type	Description
Double[]

CalculateAvgSpectrumAndVarianceSpectrumFromAmplitudeSpectrogram(Double[,])

Returns AVERAGE POWER SPECTRUM (PSD) and VARIANCE OF POWER SPECTRUM. Have been passed the amplitude spectrum but square amplitude values to get power or energy.

This method assumes that the passed amplitude spectrogram has been prepared according to method of P.D. Welch. It is the standard method used now to calculate a PSD. Welch's method splits time series into overlapping segments and windows them. It is the windowing that makes Welche's method different. Normally overlap windows because windows decay at edges and therefore loss of info. Can now do FFT. Does not need to be FFT, but if so then window must be power of 2. Square the FFT coefficients >>>> energy. Then take average in each frquncy bin. Averaging reduces the variance. Welch's method is an improvement on the standard periodogram spectrum estimating method and on Bartlett's method, in that it reduces noise in the estimated power spectra in exchange for reducing the frequency resolution. The end result is an array of power measurements vs. frequency "bin".

As well as calculating the av power spectrum, this method also returns a variance spectrum and a spectrum of the Coeff of Variation = var/mean.

Declaration

public static Tuple<double[], double[], double[]> CalculateAvgSpectrumAndVarianceSpectrumFromAmplitudeSpectrogram(double[, ] amplitudeSpectrogram)

Parameters

Type	Name	Description
Double[,]	amplitudeSpectrogram	this is an amplitude spectrum. Must square values to get power.

Returns

Type	Description
Tuple<Double[], Double[], Double[]>	three spectral indices.

CalculateAvgSpectrumFromEnergySpectrogram(Double[,])

NOTE: This method should not be used to average a decibel spectrogram. Use only for power spectrograms.

Declaration

public static double[] CalculateAvgSpectrumFromEnergySpectrogram(double[, ] spectrogram)

Parameters

Type	Name	Description
Double[,]	spectrogram

Returns

Type	Description
Double[]

CalculateNdsi(Double[], Int32, Int32, Int32, Int32)

Calculates Stuart Gage's NDSI acoustic index from the Power Spectrum derived from a spectrogram. This method assumes P.D. Welch's method has been used to calculate the PSD. See method above: CalculateAvgSpectrumAndVarianceSpectrumFromAmplitudeSpectrogram().

Declaration

public static double CalculateNdsi(double[] psd, int samplerate, int lowBound, int midBound, int topBound)

Parameters

Type	Name	Description
Double[]	psd	power spectral density.
Int32	samplerate	original sample rate of the recording. Only used to get nyquist.
Int32	lowBound	low ndsi bound.
Int32	midBound	mid ndsi bound.
Int32	topBound	top ndsi bound.

Returns

Type	Description
Double	ndsi.

CreateFalseColourAmplitudeSpectrogram(Double[,], Double[,], Byte[,])

Another experimental method to colour render spectrograms, this time amplitude spectrograms.

Declaration

public static Image<Rgb24> CreateFalseColourAmplitudeSpectrogram(double[, ] spectrogramData, double[, ] nrSpectrogramData, byte[, ] hits)

Parameters

Type	Name	Description
Double[,]	spectrogramData
Double[,]	nrSpectrogramData
Byte[,]	hits

Returns

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

CreateFalseColourDecibelSpectrogram(Double[,], Double[,], Byte[,])

TODO: This method needs a unit test. This is experimental method to explore colour rendering of standard spectrograms Used to convert a standard decibel spectrogram into a colour version using a colour rendering for three separate properties.

Declaration

public static Image<Rgb24> CreateFalseColourDecibelSpectrogram(double[, ] dbSpectrogramData, double[, ] nrSpectrogramData, byte[, ] hits)

Parameters

Type	Name	Description
Double[,]	dbSpectrogramData	the raw decibel spectrogram data - assigned to red channel.
Double[,]	nrSpectrogramData	the noise reduced decibel spectrogram data - assigned to green channel.
Byte[,]	hits	assigned to ridge colours.

Returns

Type	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>	coloured-rendered spectrogram as image.

CreateFalseColourDecibelSpectrogramForZooming(Double[,], Double[,], Byte[,])

Creates a false-coloured spectrogram from spectral frame data. That is, uses normal spectrogram data but draws the raw data in red and then superimposes the noise reduced decibel data Also uses the spectral "hits" data for highlighting the spectrogram.

IMPORTANT WARNING!!!! THIS METHOD ASSUMES THAT BOTH SPECTRAL MATRICES HAVE BEEN NORMALISED IN [0,1].

Declaration

public static Image<Rgb24> CreateFalseColourDecibelSpectrogramForZooming(double[, ] dbSpectrogramNorm, double[, ] nrSpectrogramNorm, byte[, ] hits)

Parameters

Type	Name	Description
Double[,]	dbSpectrogramNorm	the raw decibel spectrogram data - assigned to red channel.
Double[,]	nrSpectrogramNorm	the noise reduced decibel spectrogram data - assigned to green channel.
Byte[,]	hits	assigned to ridge colours.

Returns

Type	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>	coloured-rendered spectrogram as image.

DrawGridLinesOnImage(Image<Rgb24>, TimeSpan, TimeSpan, TimeSpan, FrequencyScale)

Only calls method to draw frequency lines but may in future want to add the times scale.

Declaration

public static void DrawGridLinesOnImage(Image<Rgb24> bmp, TimeSpan startOffset, TimeSpan fullDuration, TimeSpan xAxisTicInterval, FrequencyScale freqScale)

Parameters

Type	Name	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>	bmp	the spectrogram image.
TimeSpan	startOffset	start Offset.
TimeSpan	fullDuration	full Duration.
TimeSpan	xAxisTicInterval	xAxis Tic Interval.
FrequencyScale	freqScale	freq Scale.

DrawTimeLinesOnImage(Image<Rgb24>, TimeSpan, TimeSpan, TimeSpan)

Declaration

public static void DrawTimeLinesOnImage(Image<Rgb24> bmp, TimeSpan startOffset, TimeSpan fullDuration, TimeSpan xAxisTicInterval)

Parameters

Type	Name	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>	bmp
TimeSpan	startOffset
TimeSpan	fullDuration
TimeSpan	xAxisTicInterval

ExtractFreqSubband(Double[,], Int32, Int32, Boolean, Int32, Double)

Declaration

public static double[, ] ExtractFreqSubband(double[, ] m, int minHz, int maxHz, bool doMelscale, int binCount, double binWidth)

Parameters

Type	Name	Description
Double[,]	m
Int32	minHz
Int32	maxHz
Boolean	doMelscale
Int32	binCount
Double	binWidth

Returns

Type	Description
Double[,]

ExtractModalNoiseSubband(Double[], Int32, Int32, Boolean, Int32, Double)

Declaration

public static double[] ExtractModalNoiseSubband(double[] modalNoise, int minHz, int maxHz, bool doMelScale, int nyquist, double binWidth)

Parameters

Type	Name	Description
Double[]	modalNoise
Int32	minHz
Int32	maxHz
Boolean	doMelScale
Int32	nyquist
Double	binWidth

Returns

Type	Description
Double[]

FilterbankIntegral(Double[], Int32, Int32, Int32)

Implements the integral for a single filter in a filterbank.

Declaration

public static double FilterbankIntegral(double[] spectrum, int lowIndex, int centreIndex, int highIndex)

Parameters

Type	Name	Description
Double[]	spectrum	THe input (linear) spectrum.
Int32	lowIndex	The lower bound of the filter.
Int32	centreIndex	Centre index of the filter.
Int32	highIndex	Upper bound of the filter.

Returns

Type	Description
Double	The integral of the filter.

GetImage(Double[,], Int32, Boolean)

Declaration

public static Image<Rgb24> GetImage(double[, ] data, int nyquist, bool DoMel)

Parameters

Type	Name	Description
Double[,]	data
Int32	nyquist
Boolean	DoMel

Returns

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

GetImageFullyAnnotated(Image<Rgb24>, String, Int32[,], TimeSpan)

Declaration

public static Image<Rgb24> GetImageFullyAnnotated(Image<Rgb24> image, string title, int[, ] gridLineLocations, TimeSpan duration)

Parameters

Type	Name	Description
SixLabors.ImageSharp.Image<SixLabors.ImageSharp.PixelFormats.Rgb24>	image
String	title
Int32[,]	gridLineLocations
TimeSpan	duration

Returns

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

GetSonogramPlusCharts(BaseSonogram, List<AcousticEvent>, List<Plot>, Double[,], String)

This method draws a spectrogram with other useful information attached.

Declaration

public static Image<Rgb24> GetSonogramPlusCharts(BaseSonogram sonogram, List<AcousticEvent> events, List<Plot> plots, double[, ] hits, string title)

Parameters

Type	Name	Description
BaseSonogram	sonogram	of BaseSonogram class.
List<AcousticEvent>	events	a list of acoustic events.
List<Plot>	plots	a list of plots relevant to the spectrogram scores.
Double[,]	hits	not often used - can be null.
String	title

Returns

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

GetSonogramPlusCharts(BaseSonogram, List<EventCommon>, List<Plot>, Double[,], String)

This method draws a spectrogram with other useful information attached.

Declaration

public static Image<Rgb24> GetSonogramPlusCharts(BaseSonogram sonogram, List<EventCommon> events, List<Plot> plots, double[, ] hits, string title)

Parameters

Type	Name	Description
BaseSonogram	sonogram	of BaseSonogram class.
List<EventCommon>	events	a list of acoustic events.
List<Plot>	plots	a list of plots relevant to the spectrogram scores.
Double[,]	hits	not often used - can be null.
String	title

Returns

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

HistogramOfSpectralPeaks(Double[,])

Returns a HISTORGRAM OF THE DISTRIBUTION of SPECTRAL maxima.

Declaration

public static Tuple<int[], int[]> HistogramOfSpectralPeaks(double[, ] spectrogram)

Parameters

Type	Name	Description
Double[,]	spectrogram

Returns

Type	Description
Tuple<Int32[], Int32[]>

NormaliseSpectrogramMatrix(Double[,], Double, Double, Double)

Used to normalise a spectrogram in 0,1.

Declaration

public static double[, ] NormaliseSpectrogramMatrix(double[, ] matrix, double truncateMin, double truncateMax, double backgroundFilterCoeff)

Parameters

Type	Name	Description
Double[,]	matrix	the spectrogram data.
Double	truncateMin	set all values above to 1.0.
Double	truncateMax	set all values below to zero.
Double	backgroundFilterCoeff	used to de-emphisize the background.

Returns

Type	Description
Double[,]	a normalised matrix of spectrogram data.

RescaleSpectrumUsingFilterbank(Int32[,], Double[])

Converts a single linear frequency scale spectrum to a reduced linear or non-linear frequency scale spectrum. The scale conversion is defined in the transformMatrix variable. The transformMatrix defines a filter bank. Strictly speaking the input spectrum can be any vector of values but typically it should be linear spectrum.

Declaration

public static double[] RescaleSpectrumUsingFilterbank(int[, ] transformMatrix, double[] linearSpectrum)

Parameters

Type	Name	Description
Int32[,]	transformMatrix
Double[]	linearSpectrum

Returns

Type	Description
Double[]