MATLAB: Spectrogram – How to divide signal

spectrogram

Hello

I have a signal 120 seconds long with sampling frequency 1KHz. How do I compute the spectrogram with 120 time segments ? Each segment 1000 samples. The code below gives me t = 136 not 120

segments = 120;     
windowsize = 1000;
nfft = 1000;
[S,F,T,P] = spectrogram(data,windowsize,segments,nfft,Fs);

Best Answer

The third input argument of spectrogram expects the number of samples to overlap each window, not the number of segments. Please try:

   winSize = 1000;
   overlap = 0;
   fftSize = winSize;
   [S,F,T,P] = spectrogram(data,winSize,overlap,fftSize,Fs);

It is common practice, however, to set the overlap to half of the window size:

   overlap = floor(winSize/2);

but it is not required if you don't want it.

HTH.

Related Solutions

MATLAB: Code for ploting fourier transform spectrum

spectrogram()

MATLAB: 8 segments for pwelch

The use of 8 as the default value for the number of segments is purely arbitrary. It is a perfectly reasonable value, but it is no better or worse than any other choice. In general, you can often achieve better results by choosing a custom value for the window size and the overlap amount. Remember that there is a trade off between the window size and the frequency resolution. The larger the window, the smaller the frequency increment, and vice versa.

Please try:

   nSegments = ... ;
   winSize = signalLength/nSegments;
   overlap = winSize/2;
   fftSize = winSize;
   figure;
   pwelch(x,winSize,overlap,fftSize,Fs);

for a variety of values for nSegments.

Best Answer

Related Solutions

MATLAB: Code for ploting fourier transform spectrum

MATLAB: 8 segments for pwelch

Related Question