function [filt_weights,w_index] = nweights(p,pred,z)% NWEIGHTS - calculate new weights of filtering pdf%% Nonlinear Filtering Toolbox version 2.0-rc4% Copyright (c) 1995 - 2007 NFT developement Team,%              Department of Cybernetics,%              University of West Bohemia in Pilsen%***************************** Initialization ***********************%------------------- Get parameters of predictive pdf ---------------Mp             = pred.means;          % get mixture terms mean valuesVp             = pred.vars;           % get mixture terms variancesalpha_pred     = pred.weights;        % get mixture terms weightspred_terms_num = pred.n_members;      % get No. of mixture terms%---------- Calculate new number of filtering mixtue members --------filt_terms_num = pred_terms_num*p.r;alpha_filt = zeros(filt_terms_num,1); % create new weights vector%************************  Weight evaluation ************************system = p.system;                    % get estimated systemzeta   = zeros(filt_terms_num,1);     % prepare auxilinary variablefor j = 1:filt_terms_num,    i=j-fix((j-1)/pred_terms_num)*pred_terms_num;    m=1+fix((j-1)/pred_terms_num);    H = nfdiff(system.h,[Mp{i}' zeros(1, system.nv)],system.x);    h_zeta = nfeval(system.h,[Mp{i}' zeros(1, system.nv)])+p.delta*p.v_means{m};    zeta_var = H*Vp{i}*H'+p.v_vars{m};    exponent = -((z-h_zeta)'*inv(zeta_var)*(z-h_zeta))/2;    zeta(j)  = 1/(exp(system.nv*log(2*pi)/2)*sqrt(det(zeta_var)))*exp(exponent);    alpha_filt(j) = alpha_pred(i)*p.v_weights(m)*zeta(j);end%**************************** Weights sorting *****************************%pindex=zeros(filt_terms_num,1);%max_weights=zeros(filt_terms_num,1);[max_weights,pindex] = sort(-alpha_filt);correction           = length(find(abs(max_weights)<=eps));%-- return only non-zero normalized sorted weights and their index ---if p.reduction ~= -1    filt_terms_num=min(filt_terms_num-correction,p.reduction);else    filt_terms_num=filt_terms_num-correction;end;norm         = sum(max_weights(1:filt_terms_num));filt_weights = max_weights(1:filt_terms_num)/norm;w_index      = pindex(1:filt_terms_num);