How to extract peaks from a matrix and concatenate columns with different number of rows using a loop?

Hi I want to extract the peaks of all columns from a matrix M.

Each column of M should generate a new column of a matrix N

with the PEAKS as elements.

I tried the following loop:

for n=1:10

NPEAKS{n} = (findpeaks(M(:,n:n)))';

ALLPEAKS=[NPEAKS{1:n}];

end;

The second line of the loop works fine

and I obtain in each { } the desired column with the peaks.

However, the concatenation (third line of the loop)

of all columns to ONE new matrix did not work

because the amount of peaks (or elements) found for each

columns differs from column to columns.

How can I correct the commands above or to obtain the same result

doing something else.

Thank you for your help

Emerson

Answer 1

[Andrei Bobrov]

variant of using cellfun

M = randi(1500,12,12);
[p,l] = arrayfun(@(x)findpeaks(M(:,x)),1:size(M,2),'UniformOutput',false);
iend = cellfun('size',l,2);
iendmax = max(iend);
NPEAKSdouble = cell2mat(arrayfun(@(x,y,z)[cat(3,x{:}.',y{:}.');zeros(iendmax-z,1,2)],p,l,iend,'UniformOutput',false));

variant of using for loops

M = randi(1500,12,12);
nM = size(M);
N = 0;
NPEAKSdouble = zeros([nM,2]);
for jj = 1:nM(2)
  [p,l] = findpeaks(M(:,jj));
  lp = length(p);
  N = max(N,lp);
  NPEAKSdouble(1:lp,jj,:) = cat(3,p,l);
end
NPEAKSdouble = NPEAKSdouble(1:N,:,:);

[Oleg Komarov]

Going from double to cell is unnecessary when you could use arrayfun instead of cellfun.

Padding could be done using the locations by creating a zero matrix and assigning only the peaks.

[Andrei Bobrov]

Dear Oleg agree with you, I'll be corrected, hurry. Thank you