function X = randnv(mu, cx, N)
%
% function X = randnv(mu, cx, N)
%
% this function generates the normal random vector with mean mu and covariance cx.
% The function first generate the normal random vector which are indepently, 
% and then use eigen vectors to transform into multivariate gaussin random 
% vector. 
% 
% This function uses the randn() in Matlab , so the user can change or reset the
% seed of randn(). 
% 
% Output 
%  X  =  a matrix and each column is for one random variable.
% 
% Input 
% mu =  means of random variables, column vector
% cx =  covariance matrix
% N  =  number of samples
%
% example: X = randv([0;0;0], [1 0 0; 0 1 0.5; 0 0.5 1], 1500) ; 

% Copyright : Speech communciation lab, University of Maryland
%
% Written by Xinhui Zhou - April 15, 2008
% 
if nargin < 3,
	eval('help randnv');
	return;
end;
[v, lambda] = eig(cx) ; 
% randn('state',sum(100*clock)) ; % change the seed of the random generator
y = randn(N*length(mu), 1) ;
y = reshape(y, N, length(mu)) ; 
for k = 1:length(mu)
    y(:,k) = sqrt(lambda(k,k))*y(:,k)  ;
end
X = (y*v') + repmat(mu', N,1) ;
return ;