Index exceeds matrix dimensions

I run a program of Improved Dense Trajectory features Encoded by Fisher Vector, written by Bo Yang. Below shows a part of the coding.

for i=1:dtf_feat_num
	feat=feats_train{i};
	
	% L1 normalization & Square root
	feat=sqrt(feat/norm(feat,1));
	
	% Do PCA on train/test data to half-size original descriptors
	fprintf('Doing PCA ...\n');
	pca_coeff{i} = princomp(feat');
	pca_coeff{i} = pca_coeff{i}(:, 1:floor(size(feat,1)/2))';
	% dimensionality reduction
	feat = pca_coeff{i} * feat;

	fprintf('Training Guassian Mixture Model ...\n');
	gmm{i}=gmm_gen_codebook(feat,gmm_params);
end

When i run the master.m , it shows the error below:

Index exceeds matrix dimensions.

Error in pretrain (line 57)
	pca_coeff{i} = pca_coeff{i}(:, 1:floor(size(feat,1)/2))';

Can anyone help me to solve this problem? Thank you.

[Geoff Hayes]

Mohamad - what are the dimensions of pca_coeff? When this error occurs, what are the values of i and 1:floor(size(feat,1)/2)?

[Geoff Hayes]

Mohamad's answer moved here

The dimension of pca_coeff is 4x1.

The code below are the looping part:

for i=1:dtf_feat_num
	feat=feats_train{i};
	
	% L1 normalization & Square root
	feat=sqrt(feat/norm(feat,1));
	
	% Do PCA on train/test data to half-size original descriptors
	fprintf('Doing PCA ...\n');
	pca_coeff{i} = pca(feat');
	pca_coeff{i} = pca_coeff{i}(:, 1:floor(size(feat,1)/2))';
	% dimensionality reduction
	feat = pca_coeff{i} * feat;

	fprintf('Training Guassian Mixture Model ...\n');
	gmm{i}=gmm_gen_codebook(feat,gmm_params);
end

end

Below are value of the variables in the coding:

dtf_feat_num = 4
feat = 96x985 double
floor(size(feat,1)/2) = 48

[Geoff Hayes]

And what are the dimensions of the elements in the pca_coeff array? You may need to put a breakpoint at this line, run your code and step through it with the debugger to see what is happening.