matlab mtm,Thomson multitaper spectrum

spectrum.mtm

Thomson multitaper spectrum

Syntax

Hs = spectrum.mtm

Hs = spectrum.mtm(TimeBW)

Hs = spectrum.mtm(DPSS,Concentrations)

Hs = spectrum.mtm(...,CombineMethod)

Description

Note

The use of spectrum.mtm is not recommended.

Use pmtm instead.

Hs = spectrum.mtm returns

a default Thomson multitaper spectrum object, Hs that

defines the parameters for the Thomson multitaper spectral estimation

algorithm, which uses a linear or nonlinear combination of modified

periodograms. The periodograms are computed using a sequence of orthogonal

tapers (windows in the frequency domain) specified from discrete prolate

spheroidal sequences (dpss).

This object uses the following default values:

Property NameDefault ValueDescriptionTimeBW4Product of time and bandwidth for the discrete prolate

spheroidal sequences (or Slepian sequences) used as data windows

CombineMethod'adaptive'Algorithm for combining the individual spectral estimates.

Valid values are 'adaptive' — adaptive

(nonlinear) 'unity' — unity weights (linear) 'eigenvector' —

Eigenvalue weights (linear)

Hs = spectrum.mtm(TimeBW) returns

a spectrum object, Hs with the specified time-bandwidth

product.

Hs = spectrum.mtm(DPSS,Concentrations) returns

a spectrum object, Hs with the specified dpss data

tapers and their concentrations.

Note

You can either specify the time-bandwidth product (TimeBW)

or the DPSS data tapers and their Concentrations.

See dpss and pmtm for more information.

Hs = spectrum.mtm(...,CombineMethod) returns

a spectrum object, Hs, with the specified method

for combining the spectral estimates. Refer to the table above for

valid CombineMethod values.

Examples

Define a cosine of 200 Hz, add noise and view its power spectral

density using the Thomson multitaper algorithm with a time-bandwidth

product of 3.5.

Fs=1000;

t=0:1/Fs:.3;

x=cos(2*pi*t*200)+randn(size(t));

Hs=spectrum.mtm(3.5);

psd(Hs,x,'Fs',Fs)

The above example could be done by specifying the data tapers

and concentrations instead of the time-bandwidth product.

Fs=1000;

t=0:1/Fs:.3;

x=cos(2*pi*t*200)+randn(size(t));

[e,v]=dpss(length(x),3.5);

Hs=spectrum.mtm(e,v);

psd(Hs,x,'Fs',Fs)

Introduced before R2006a