Signal auto-correlation in matlab (xcorr function)

Question

Panos Kerezoudis 2023-4-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1950483-signal-auto-correlation-in-matlab-xcorr-function

评论： Panos Kerezoudis 2023-4-21

Hi!

I am working on an EEG-electrophysiology dataset ('csd'). It is structured as a 3D array (channels x time x trials). The dimensions are 16x1527x200.

I am trying to calculate trial-averaged autocorrelation for each channel and I am getting the error message below. I think it has to do with the last part of my loop? The dimensions of the ac0 output (dot products) are 1x201 for some reason, I think the way the function is built, it adds one more column?

Any help would be great!

2 个评论
显示无隐藏无

Mathieu NOE 2023-4-20

if you expect someone to help you , it's good practice to share some code and data

Panos Kerezoudis 2023-4-20

在 MATLAB Online 中打开

Absolutely. The code is below. I cannot share data unfortunately (confidentiality reasons).

% Calculate trial-averaged auto-correlation
nchan = size(csd, 1); % channels
ntrials = size(csd, 3); % trials
nlags = 100; % n of lags 
ac = zeros(1, 2*nlags + 1); % initialize auto-correlation vector
for i = 1:nchan % loop over channels
    for k = 1:ntrials % loop over trials
    x = csd(i, :, k) - mean(csd(i, :, k)); % mean-normalize data
    [acO, lags] = xcorr(x, nlags, 'biased'); % biased estimate
    ac(i, :) = ac(i, :) + double(acO(i, :)/ntrials); % add to total scaled by 1/k
    end
end

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Chunru 2023-4-21

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1950483-signal-auto-correlation-in-matlab-xcorr-function#answer_1220008

在 MATLAB Online 中打开

csd = randn(4, 100, 20);        % dummy data
% Calculate trial-averaged auto-correlation
nchan = size(csd, 1); % channels
ntrials = size(csd, 3); % trials
nlags = 100; % n of lags 
ac = zeros(nchan, 2*nlags + 1); % initialize auto-correlation vector
%          ^^^^
for i = 1:nchan % loop over channels
    for k = 1:ntrials % loop over trials
        x = csd(i, :, k) - mean(csd(i, :, k)); % mean-normalize data
        [acO, lags] = xcorr(x, nlags, 'biased'); % biased estimate
        
        ac(i, :) = ac(i, :) + double(acO/ntrials); % add to total scaled by 1/k
        %                            ^^^
    end
end