spectrogram of a matrix with 32 channels in one graph with color

Hi all,

I have matrix of EEG data with 32 channels.

I want to create a stacked spectrogram for all 32 channels in one graph to find the hot spots.

I like the x-axis to show the frequency, the y-axis all my channels stacked on one another.

I really appreciate your help.

Answer 1

[Chunru]

% generate some data
fs = 500;
x = sin(2*pi*(0:1023)'*(1:32)/fs) + randn(1024, 32);


for i=1:size(x, 2)
    [s,f,t] = spectrogram(x(:,i), 256,128,256, fs, 'psd');
    if i==1
        idx  = find(f<=55);        % get the freq range
        nf = length(idx);
        faxis = f(idx);
        
        sall = zeros(nf, size(s,2), 32);
    end
    sall(:, :, i) = 20*log10(abs(s(idx, :)));
end

imagesc(faxis, [1:32], reshape(sall, nf, [])');
xlabel('f (Hz)');
ylabel('Ch')

[Fatemeh Yousef Zadeh]

Hi again

thanks for your answer.

To clarify: my original sampling frequency was 500 but I am only interested to see frequencies up to 55 to find the hot spots. Should I modify the code as below:

fs=500,

[s,f,t] = spectrogram(x(:,i), 1000,500,1000, fs, 'psd');

thanks again

[Chunru]

see update above

[Fatemeh Yousef Zadeh]

thanks

[Prajakta Rathod]

This answer is very much helpful for me. I just want to know what this color indicates. power or amplitude ?

[Chunru]

[s,f,t] = spectrogram(x(:,i), 256,128,256, fs, 'psd'); return s as short-time-fouriour-transform. 20*log10(abs(s)) is spectrum density estimate (subject to a normalization factor of NFFT and sampling frequency).

if you want to get the ower spectral density or power spectrum direction, you cab use [s,f,t, ps] = spectrogram(x(:,i), 256,128,256, fs, 'psd'); Here ps is the one-sided (for real signal) modified periodogram (a power spectral density estimate).

To recap, the plot above show power spectrum in dB (subject to a constant factor)

doc spectrogram for more details.

[Prajakta Rathod]

Sir, thank you very much for your reply.