function q = importance_weights(xuw,xud,Xs_fft,id_v,delta,xuvar);

xu=xuw;

% X=xu(3,:);
% xu_thite=xu(1,:);
% xu_Qv=xu(2,:);
% xu_fai=xu(3,:);
% 
% mu_thite=mean(xu_thite)+xu_thite;
% mu_Qv=mean(xu_Qv)+xu_Qv;
% mu_fai=mean(xu_fai)+xu_fai;
% 
% sigma_thite=var(xu_thite);
% sigma_
[rows,samnum]=size(xu);
X = sh_max(xuw,xud,Xs_fft,id_v,delta,xuvar);
Qm = mean(xu.').'+X;
sigma=var(xu.').';

q=[normrnd(Qm(1),sigma(1),1,samnum);normrnd(Qm(1),sigma(1),2,samnum);normrnd(Qm(1),sigma(1),3,samnum)];
q_sum=sum(q,2);
q=[q(1,:)./q_sum(1);q(2,:)./q_sum(2);q(3,:)./q_sum(3)];
q=abs(q);
% 

% % PURPOSE : Computes the normalised importance ratios for the 
% %           model described in the file sirdemo1.m.
% % INPUTS  : - xu = The predicted state samples.
% %           - y = The output measurements.
% %           - R = The measurement noise covariance.
% % OUTPUTS : - q = The normalised importance ratios.
% 
% % AUTHOR  : Nando de Freitas - Thanks for the acknowledgement :-)
% % DATE    : 08-09-98
% 
% 
% if nargin < 3, error('Not enough input arguments.'); end
% 
% [rows,cols] = size(xu);
% q = zeros(size(xu));
% m = (xu.^(2))./20;
% for s=1:rows,
%   q(s,1) = exp(-.5*R^(-1)*(y- m(s,1))^(2))./sum(exp(-.5*R^(-1)*(y.*ones(size(xu))-m).^(2)));
% end;
% q_sum=sum(q);
% q = q./q_sum;   
% subplot(224); 
% plot(xu,q,'+')      
% ylabel('Likelihood function','fontsize',15);
% xlabel('Hidden state support','fontsize',15)
% axis([-30 30 0 0.03]);
%