MATLAB: How to specify overlap in dsp.SpectrumEstimator object

DSP System Toolboxdsp.spectrumestimatoroverlappwelch

The 'pwelch' function has an option to specify the overlap between the divided segments of the signal whose spectral estimate is to be calculated. However, its equivalent object 'dsp.SpectrumEstimator' has no option for 'overlap' property. How do I specify overlap in 'dsp.SpectrumEstimator' object?

Best Answer

Overlap property does not exist in the 'dsp.SpectrumEstimator' object as of MATLAB R2017a. To implement the overlap option there are two options -

1. Use the 'dsp.AsyncBuffer' object to capture the overlapped frame and then step that through the object 'dsp.SpectrumEstimator'. This object was introduced in R2017a.

2. Using vector indexing, manually divide the input signal into frames including the overlapping samples. Then step these frames through the 'dsp.SpectrumEstimator', one at a time.

Please refer to the attached script which shows the use of 'dsp.AsyncBuffer' object to implement the overlap in 'dsp.SpectrumEstimator' and obtain the equivalent spectrum as obtained from 'pwelch'.

Related Solutions

MATLAB: Does Multi Channel Audio with the MOTU UltraLite-mk3 does not work (audioDeviceWriter)

The correct syntax for this workflow is the following:

deviceWriter = audioDeviceWriter('SampleRate',Fs,'BitDepth','16-bit integer',...
'ChannelMappingSource',"Property",'ChannelMapping',1:8,'BufferSize',1024,...
'Driver','ASIO','Device','MOTU Audio ASIO');

Please use ASIO for multi-channel audio for this device.

MATLAB: Do results differ between pwelch and dsp.SpectrumEstimator

The 'window' property in 'pwelch' function specifies the length of segments which the signal gets divided into. Power spectrum is then calculated by taking the spectral averages of these segments.

The 'dsp.SpectrumEstimator' object has a property 'Spectral Averages' which determines the number of spectral averages to be taken for the segmented signal. However, the object itself does not divide the signal into segments. Although, the underlying implementation of this object is through Welch Spectrum Estimate algorithm which requires the signal to be segmented, we have to manually implement the segmentation part of the signal and step these segments through the object.

Please refer to the attached script which shows an example of correct usage of 'dsp.SpectrumEstimator' for computing the Welch Spectrum of a sinusoid signal. Please note, the parameters such as sampling frequency, length of fft, etc. are defined in the script.

Best Answer

Related Solutions

MATLAB: Does Multi Channel Audio with the MOTU UltraLite-mk3 does not work (audioDeviceWriter)

MATLAB: Do results differ between pwelch and dsp.SpectrumEstimator

Related Question