Zero pad vectors within cell array to make them equal length

I have a cell array with vectors y1, y2, y3, y4, y5 which are of variable lengths. I get the maximum length among all elements in the cell array using:

sigvecarray = {y1, y2, y3, y4, y5};
[maxsamples, idx] = max(cellfun(@numel, sigvecarray));

Now, I need to zero pad the shorter vectors with the differrence from maxsmaples and their own sample numbers.

I tried the following first:

signalvectors = {}
for k = 1:numel(sigvecarray)
    currveclength = length(sigvecarray{k})
    if currveclength < maxsamples
        padding = samples - currveclength
        signalvectors{k} = [sigvecarray{k}, zeros(padding, 1)]
    end
end

It gave me the follwing error:

Error using horzcat
Dimensions of matrices being concatenated are not consistent.

Then I tried the solution from the post here as follows:

origsamplesarray = cellfun(@numel, sigvecarray);
padfun = @(k) [sigvecarray{k} zeros(maxsamples(k) - origsamplesarray(k), 1)] ;
signalvectors = arrayfun(padfun, 1:numel(sigvecarray) , 'un', 0);

It gave me the following error:

Error using horzcat
Dimensions of matrices being concatenated are not consistent.

Error in xcorr>@(k)[sigvecarray{k},zeros(maxsamples(k)-origsamplesarray(k),1)]

What am I possibly doing wrong above?

Answer 1

[Star Strider]

I do not know what the dimensions of ‘sigvecarray’ are, however I assume they are all column vectors.

If so, replace the (,) with a (;) here, and it should work (unless there are also other problems):

signalvectors{k} = [sigvecarray{k}; zeros(padding, 1)]

EDIT — (9 Oct 2020 at 12:34)

All of these — including this:

padfun = @(k) [sigvecarray{k} zeros(maxsamples(k) - origsamplesarray(k), 1)] ;

function — are concatenating something with a column vector, that column vector being defined as:

zeros(maxsamples(k) - origsamplesarray(k), 1)

If ‘sigvecarray{k}’ is a row vector or even another column vector of a different length, that error will be thrown, since (,) is a horizontal concatenation operator.

If ‘sigvecarray{k}’ is a column vector (and that is still unknown), concatenate them vertically — not horizontally — using the (;) vertical concatenation operator:

padfun = @(k) [sigvecarray{k}; zeros(maxsamples(k) - origsamplesarray(k), 1)] ;
                             ↑ ← VERTICAL CONCATENATION OPERATOR

.

[skrowten_hermit]

Thanks a lot. I feel silly. This is a classic yet longer method I knew would work.

Any idea what is going wrong with the following:

origsamplesarray = cellfun(@numel, sigvecarray);
padfun = @(k) [sigvecarray{k} zeros(maxsamples(k) - origsamplesarray(k), 1)] ;
signalvectors = arrayfun(padfun, 1:numel(sigvecarray) , 'un', 0);

[skrowten_hermit]

Sorry, I forgot to mention with a proper edit. 'sigvecarray' is a cell array with dimensions 1x8 holding vectors (of size between 64000-72000) in each of 8 columns. I made the following changes after taking care of the error pointed out by you (the error was that I had mistaken added 'maxsamples(k)' instead of 'maxsamples':

origsamplesarray = cellfun(@numel, sigvecarray);
padfun = @(k) [sigvecarray{k}; zeros(maxsamples - origsamplesarray(k), 1)] ;
signalvectors = arrayfun(padfun, 1:numel(sigvecarray) , 'un', 0);

for k = 1:numel(signalvectors)
    disp(length(signalvectors{k}))
end

And I see that the new 'signalvectors' has all cells with vectors of size 'maxsamples'.

[Star Strider]

The only change I would suggest is to replace the arrrayfun call with a simple for loop. The arrayfun function is fine for some purposes, however it is much slower and less efficient than a for loop. With large amounts of data, that ineffficiency can become significant.

[skrowten_hermit]

Point noted! Yeah. Mine has large amounts of data. Will make the changes as you have suggested. Thanks.

[Star Strider]

As always, my pleasure!