For loop for applying filter for each column

Hi I've a signal which I read into matlab as a 4096x28 matrix. I want to apply for loop as follows which will read one column apply the filter and then store the values and then moves to the next column.

X = reshape(T_hor,[],28); % forming 4096x28 matrix 
 for i = 1:28;
  Xnew = X(:,i) - mean(X(:,i)); % filter for DC offset
 end;

It just reads the last column (28th) and stops. Is something wrong with my code? Can someone help me to fix this. Thanks in advance.

thanks in advance

Answer 1

[Greig]

Xnew is not indexed. You should have Xnew(:, i).

You could do this without the loop. One such way would be...

Xnew = X - repmat(mean(X), 4096, 1);

Another is....

Xnew = detrend(X,0);

I would go with the second (less typing), but either will save you some time in a loop.

[Bharath]

Thanks for your answer. I still have the peaks at the 0 frequency when I use the 'detrend' for the complete signal. Its getting reduced when I apply the filter for individual columns. I don't know why. Looks quiet strange. Any possible alternative ways of doing it?

[Bharath]

I get this error when I even Xnew is indexed ??? Attempted to access X(:,2); index out of bounds because size(X)=[4096,1].

[Ilham Hardy]

The answer is on the error message, you want to put data into second column of X, hence X(:,2) , which does not exist since the size of X is 4096x1 (array, not matrix/table).

[Bharath]

Thanks for your answer. But my initial matrix was of size 4096x28. And then I defined for loop. But Now I guess my for loop is overwriting the 'X' matrix. How to avoid that. I mean read every column of X and apply the filters and store them in Xnew columnwise.

[Greig]

If you really insist on using a loop try this....

X=rand(4096, 28); % My test data
Xnew = NaN(size(X)); % Preallocate for speed
for ii=1:size(X, 2)
Xnew(:,ii) = X(:,ii) - mean(X(:,ii));
end

[It is recommended not to use "i" or "j" as an indices given that they are sqrt(-1).]

Then compare with these...

Xnew2 = X - repmat(mean(X), size(X, 1), 1);
Xnew3 = detrend(X,0);

Of course if you want to apply more complicated filters than mean subtraction, the loop version may be the better option.

[Bharath]

Thanks a ton. It was very informative and it works perfectly..! :) So is it always good to preallocate before start of a loop? In my previous case I just defined as

   ii= 1:28

It just read only the last column but now it seems working perfectly good with the preallocation.

[Greig]

Your loop read through all of the columns, but since Xnew was overwritten each time, it only remembered the last loop.