setExtractorParameters

Set nondefault parameter values for feature extractor object

Since R2021b

The function has changed. For more information, see Version History.

Syntax

setExtractorParameters(freqFE,freqFeatureName,freqFeatureParams)

setExtractorParameters(tfFE,tfAnalysisName,tfAnalysisParams)

setExtractorParameters(tfFE,tfFeatureName,tfFeatureParams)

Description

setExtractorParameters(freqFE,freqFeatureName,freqFeatureParams) sets the parameters specified in freqFeatureParams to the frequency-domain feature specified in freqFeatureName. The feature extractor object freqFE uses the specified parameters to set up feature extraction.

setExtractorParameters(tfFE,tfAnalysisName,tfAnalysisParams) sets the parameters specified in tfAnalysisParams to the time-frequency analysis method specified in tfAnalysisName. The feature extractor object tfFE uses the specified parameters to set up feature extraction.

setExtractorParameters(tfFE,tfFeatureName,tfFeatureParams) sets the parameters specified in tfFeatureParams to the time-frequency feature specified in tfFeatureName. The feature extractor object tfFE uses the specified parameters to set up feature extraction.

Examples

Set Parameter Values for Feature Extraction

Create a signalFrequencyFeatureExtractor object to extract the Welch power spectral density (PSD) estimate of a signal consisting of a 100 Hz sinusoid in additive N(0,1) white noise. The sample rate is 1 kHz and the signal has a duration of 5 seconds.

Fs = 1000;
t = 0:1/Fs:5-1/Fs;
x = cos(2*pi*100*t) + randn(size(t));

sFE = signalFrequencyFeatureExtractor(SampleRate=Fs,WelchPSD=true);

For the PSD computation, set the OverlapLength to 25 samples and the FFTLength to 512 samples. Call the getExtractorParameters function on the object to view the PSD parameters.

setExtractorParameters(sFE,"WelchPSD",OverlapLength=25,FFTLength=1024)
params = getExtractorParameters(sFE,"WelchPSD")

params = struct with fields:
          FFTLength: 1024
    FrequencyVector: []
      OverlapLength: 25
             Window: []

Use the extract function to extract the Welch's PSD estimate of the signal. Plot the result.

psdFeature = extract(sFE,x);

f = linspace(0,Fs/2,length(psdFeature));
plot(f,pow2db(abs(psdFeature)))

xlabel("Frequency (Hz)")
ylabel("PSD (dB/Hz)")
grid

Input Arguments

Frequency Domain

freqFE — Frequency feature extractor

signalFrequencyFeatureExtractor object

Frequency feature extractor, specified as a signalFrequencyFeatureExtractor object.

freqFeatureName — Name of frequency feature

"OccupiedBandwidth" | "PowerBandwidth" | "PeakAmplitude" | "PeakLocation" | "WelchPSD"

Name of frequency feature, specified as one of these values:

• "OccupiedBandwidth" — Occupied bandwidth

• "PowerBandwidth" — Power bandwidth

• "PeakAmplitude" — Amplitudes of spectral peaks

• "PeakLocation" — Locations of spectral peaks

• "WelchPSD" — Welch's power spectral density estimate

Data Types: char | string

freqFeatureParams — Parameters for frequency feature

name-value arguments | struct array

Parameters for frequency feature, specified as name-value arguments or as a struct array. Use this argument to set the parameters to extract the feature specified in freqFeatureName.

This table lists the parameters that you can set for frequency-domain signal features. You can specify multiple parameters for a feature at a time. The order of the specified parameters does not matter.

Feature Name	Parameter Name	Parameter Value
"OccupiedBandwidth"	Percentage — Power percentage	99 (default) | positive scalar less than 100
	For more information, see obw.
"PowerBandwidth"	RelativeAmplitude — Power level drop, in dB	10*log10(2) (default) | real-valued scalar
	For more information, see powerbw.
"PeakAmplitude" "PeakLocation"	PeakType — Type of peak	"maxima" (default) | "minima"
	MaxNumExtrema — Maximum number of peaks	1 (default) | positive integer scalar
	MinProminence — Minimum prominence	0 (default) | nonnegative scalar
	MinSeparation — Minimum peak separation	0 (default) | nonnegative scalar
	FlatSelection — Element of a flat region to indicate as peak	"center" (default) | "first" | "last" | "all"
	For more information about these parameters, see islocalmax or islocalmin.
"WelchPSD"	Window — Window to partition input signal into segments	Integer scalar | vector
	OverlapLength — Number of overlapping samples	Positive integer scalar
	FFTLength — Number of DFT points	Positive integer scalar
	FrequencyVector — Frequencies at which to estimate PSD	Real-valued vector FrequencyVector cannot be specified simultaneously with FFTLength.
	For more information about these parameters, see pwelch.

Example: setExtractorParameters(freqFE,"PeakAmplitude",PeakType="maxima",MaxNumExtrema=2) specifies the peak type and the maximum number of extrema parameters for the PeakAmplitude feature using name-value arguments.

Example: PeakAmplitudeParams = struct(PeakType="maxima",MaxNumExtrema=2); setExtractorParameters(freqFE,"PeakAmplitude",PeakAmplitudeParams) specifies the peak type and the maximum number of extrema parameters for the PeakAmplitude feature using a struct array.

Data Types: string | struct

Time-Frequency Domain

tfFE — Time-frequency feature extractor

signalTimeFrequencyFeatureExtractor object

Time-frequency feature extractor, specified as a signalTimeFrequencyFeatureExtractor object.

tfAnalysisName — Name of time-frequency analysis method

"spectrogram" | "synchrosqueezedspectrogram" | "emd" | "vmd" | "scalogram" | ...

Name of time-frequency analysis method, specified as one of these values:

• "spectrogram" — Short-time Fourier transform

• "synchrosqueezedspectrogram" — Fourier synchrosqueezed transform

• "emd" — Empirical mode decomposition

• "vmd" — Variational mode decomposition

• "scalogram" — Continuous wavelet transform magnitude

• "synchrosqueezedscalogram" — Wavelet synchrosqueezed transform

• "wavelet" — Maximal overlap discrete wavelet transform

• "waveletpacket" — Maximal overlap discrete wavelet packet transform

Data Types: char | string

tfAnalysisParams — Parameters for time-frequency analysis method

name-value arguments | struct array

Parameters for time-frequency analysis method, specified as name-value arguments or as a struct array. Use this argument to set the parameters to extract the feature specified in tfAnalysisName.

This table lists the parameters that you can set for time-frequency analysis methods. You can specify multiple parameters for an analysis method at a time. The order of the specified parameters does not matter.

Feature Name	Parameter Name	Parameter Value
"spectrogram"	Leakage — Spectral leakage	0.5 (default) | scalar between 0 and 1
	OverlapPercent — Overlap between adjoining segments	Positive scalar less than 100
	TimeResolution — Time resolution	Scalar greater than 1 and less than signal duration
	For more information about these parameters, see pspectrum.
"synchrosqueezedspectrogram"	Window — Window for partitioning input signal into segments	Vector of nonnegative scalars
	For more information, see fsst.
"emd"	Interpolation — Interpolation method to construct envelope	"spline" (default) — For smooth input signal | "pchip" — For nonsmooth input signal
	MaxEnergyRatio — Signal to residual energy ratio	20 (default) | real-valued scalar
	MaxNumExtrema — Maximum number of extrema in residual signal	1 (default) | positive integer scalar
	MaxNumIMF — Maximum number of intrinsic mode functions	10 (default) | positive integer scalar
	For more information about these parameters, see emd.
"vmd"	CentralFrequencies — Initial central IMF frequencies	Vector of NumIMFs positive scalars
	InitializeMethod — Method to initialize central frequencies	"peaks" (default) — Peak frequencies of the signal in frequency domain | "random" — Random frequencies uniformly distributed within [0, 0.5] cycles/sample interval | "grid" — Uniformly sampled grid of frequencies from 0 to 0.5 cycles/sample
	NumIMFs — Number of intrinsic mode functions	5 (default) | positive integer scalar
	PenaltyFactor — Penalty factor	1000 (default) | positive scalar
	For more information about these parameters, see vmd.
"scalogram"	Boundary — Signal extension boundary	"reflection" (default) — Extend signal symmetrically at the left and right boundaries | "periodic" — Extend signal periodically | "zeropad" — Pad signal boundaries with zeros
	Frequencylimits — Frequency limits	Two-element row vector of real-valued scalars
	VoicesPerOctave — Number of voices per octave	10 (default) | integer scalar between 1 and 48
	Wavelet — Name of analytic wavelet	"morse" (default) — Analytic Morse wavelet | "amor" — Analytic Morlet wavelet | "bump" — Bump wavelet
	WaveletParameters — Symmetry and time-bandwidth product of the Morse wavelet	[3 60] (default) | two-element vector
	For more information about these parameters, see cwt (Wavelet Toolbox).
"synchrosqueezedscalogram"	Boundary — Signal extension boundary	"reflection" (default) — Extend signal symmetrically at the left and right boundaries | "periodic" — Extend signal periodically | "zeropad" — Pad signal boundaries with zeros
	VoicesPerOctave — Number of voices per octave	32 (default) | even integer scalar between 10 and 48
	Wavelet — Name of analytic wavelet	"morse" (default) — Analytic Morse wavelet | "amor" — Analytic Morlet wavelet | "bump" — Bump wavelet
	For more information about these parameters, see wsst (Wavelet Toolbox).
"wavelet"	HighPass — Wavelet (highpass) filter LowPass — Scaling (lowpass) filter	Even-length real-valued vectors You must specify LowPass and HighPass simultaneously and with the same length. Do not specify these parameters if you specify Wname.
	Level — Transform level	Positive integer scalar less than or equal to floor(log2(N)), where N is the signal length.
	Reflection — Option to use reflection boundary handling	strings(0) (default) | "reflection"
	Wname — Wavelet name	"sym4" (default) | character vector | string scalar
	For more information about these parameters, see modwt (Wavelet Toolbox).
"waveletpacket"	FullTree — Option to return full wavelet packet tree	false or 0 (default) | true or 1
	HighPass — Wavelet (highpass) filter LowPass — Scaling (lowpass) filter	Even-length real-valued vectors You must specify LowPass and HighPass simultaneously and with vectors of the same length. Do not specify these parameters if you specify Wname.
	Level — Transform level	Positive integer scalar less than or equal to floor(log2(N)), where N is the signal length.
	Wname — Wavelet name	"fk18" (default) | character vector | string scalar
	For more information about these parameters, see modwpt (Wavelet Toolbox).

Example: setExtractorParameters(tfFE,"emd",Interpolation="spline",MaxNumExtrema=2) specifies the interpolation method and the maximum number of extrema parameters for the empirical mode decomposition (EMD) analysis method using name-value arguments.

Example: emdParams = struct(Interpolation="spline",MaxNumExtrema=2); setExtractorParameters(tfFE,"emd",emdParams) specifies the interpolation method and the maximum number of extrema parameters for the EMD analysis method using a struct array.

Data Types: string | struct

tfFeatureName — Name of time-frequency feature

"SpectralEntropy" | "TFRidges" | "InstantaneousBandwidth" | "InstantaneousFrequency" | "InstantaneousEnergy" | ...

Name of time-frequency feature, specified as one of these values:

• "SpectralEntropy" — Spectral entropy

• "TFRidges" — Time-frequency ridges

• "InstantaneousBandwidth" — Instantaneous bandwidth

• "InstantaneousFrequency" — Instantaneous frequency

• "InstantaneousEnergy" — Instantaneous energy

• "WaveletEntropy" — Wavelet entropy

• "TimeSpectrum" — Time spectrum

• "ScaleSpectrum" — Scale spectrum

Data Types: char | string

tfFeatureParams — Parameters for time-frequency feature

name-value arguments | struct array

Parameters for time-frequency feature, specified as name-value arguments or as a struct array. Use this argument to set the parameters to extract the feature specified in tfFeatureName.

This table lists the parameters that you can set for time-frequency signal features. You can specify multiple parameters for a feature at a time. The order of the specified parameters does not matter.

Feature Name	Parameter Name	Parameter Value
"SpectralEntropy"	Range — Frequency range, in Hz	Two-element row vector of nonnegative scalars
	For more information, see spectralEntropy.
"TFRidges"	NumRidges — Number of time-frequency ridges	1 (default) | Positive integer scalar
	NumFrequencyBins — Number of bins to remove when extracting multiple ridges	4 (default) | Positive integer scalar
	Penalty — Penalty for changing frequency	Nonnegative real-valued scalar
	For more information about these parameters, see tfridge.
"InstantaneousBandwidth"	FrequencyLimits — Frequency range, in Hz	Two-element row vector of real-valued scalars
	ScaleFactor — Scaling factor for spectral moment	sqrt(4*pi) (default) | Positive scalar
	For more information about these parameters, see instbw.
"InstantaneousFrequency" Note This feature supports the FrequencyResolution and MinThreshold parameters for these analysis methods: "emd", "vmd", "wavelet", and "waveletpacket".	FrequencyLimits — Frequency range, in Hz	Two-element row vector of real-valued scalars
	FrequencyResolution — Frequency resolution to discretize frequency limits	-Inf (default) | scalar
	MinThreshold — Minimum threshold value of Hilbert spectrum	Real-valued scalar
	For more information, see instfreq or hht.
"InstantaneousEnergy" Note This feature supports extraction parameters for these analysis methods: "emd", "vmd", "wavelet", and "waveletpacket".	FrequencyLimits — Frequency limits to compute Hilbert spectrum	Two-element row vector of real-valued scalars
	FrequencyResolution — Frequency resolution to discretize frequency limits	-Inf (default) | scalar
	MinThreshold — Minimum threshold value of Hilbert spectrum	Real-valued scalar
	For more information about these parameters, see hht.
"WaveletEntropy"	Entropy — Entropy measure	"Shannon" (default) | "Renyi" | "Tsallis"
	Exponent — Exponent for Renyi and Tsallis entropy	2 (default) | Real-valued scalar
	Distribution — Normalization method for empirical probability distribution and wavelet transform coefficients	"scale" (default) — Normalize the squared magnitudes of the coefficients by their total sum | "global" — Normalize the wavelet coefficients at each scale
	Scaled — Option to scale wavelet entropy	true or 1 (default) | false or 0
	For more information about these parameters, see wentropy (Wavelet Toolbox).
"TimeSpectrum"	Normalization — Normalization method for time-averaged wavelet spectrum	"var" (default) — Normalize to equal the variance of the input signal | "pdf" — Normalize to equal 1 | "none" — Do not normalize
	SpectrumType — Type of wavelet spectrum	"power" (default) — Normalize the averaged sum of the time-averaged wavelet spectrum | "density" — Normalize the weighted integral of the wavelet spectrum
	TimeLimits — Time limits for averaging wavelet spectrum	Two-element vector of real-valued scalars
	For more information about these parameters, see timeSpectrum (Wavelet Toolbox).
"ScaleSpectrum"	Normalization — Normalization method for time-averaged wavelet spectrum	"var" (default) — Normalize to equal the variance of the input signal | "pdf" — Normalize to equal 1 | "none" — Do not normalize
	SpectrumType — Type of wavelet spectrum	"power" (default) — Normalize the averaged sum of the time-averaged wavelet spectrum | "density" — Normalize the weighted integral of the wavelet spectrum
	FrequencyLimits — Frequency limits for averaging magnitude-squared scalogram	Two-element vector of real-valued scalars
	For more information about these parameters, see scaleSpectrum (Wavelet Toolbox).

Example: setExtractorParameters(tfFE,"TFRidges",NumRidges=3,NumFrequencyBins=10) specifies the number of ridges and number of frequency bins parameters for the TFRidges feature using name-value arguments.

Example: tfRidgesParams = struct(NumRidges=3,NumFrequencyBins=10); setExtractorParameters(tfFE,"TFRidges",tfRidgesParams) specifies the number of ridges and number of frequency bins parameters for the TFRidges feature using a struct array.

Data Types: string | struct

Tips

• Use the getExtractorParameters function to get the list of available parameters to use when extracting the features that you enabled when creating the feature extractor object.

Version History

Introduced in R2021b

R2026a: The setExtractorParameters function has changed

The setExtractorParameters function introduces support for the Boundary extraction parameter for the "scalogram" and "synchrosqueezedscalogram" time-frequency analysis methods.

For the "synchrosqueezedscalogram" analysis method, the Boundary parameter replaces the ExtendSignal parameter. This table list the updates that you must make to your code.

Analysis Method	Name-Value Argument in R2025b or Earlier	Updated Name-Value Argument in R2026a
"synchrosqueezedscalogram"	ExtendSignal=0	Boundary="periodic"
	ExtendSignal=1	Boundary="reflection"
	N/A	Boundary="zeropad"