function new_feature = normal(feature)
% NORMAL Data normalization for pattern recognition
%	Usage: new_feature = normal(feature)
%	where "feature" is the feature matrix of the design set, with
%	each column being the values for a specific feature.
%	The function "normal" returns a new feature matrix "new_feature"
%	where each column is a set of sample data of Gaussian
%	distribution with zero mean and unity variance.

sample_n = size(feature, 1);
mean_vec = mean(feature);
std_vec = sqrt(diag(cov(feature)));
new_feature = feature-ones(sample_n,1)*mean_vec;
new_feature = new_feature*diag(1./std_vec);