normdis.m

模式识别中Parzen窗函数的实现

%two-dimensional Normal Distribution
%
% (C)2008 TangSheng
%
function NormDis(u1,u2,sig1,sig2,rho)

if nargin < 5, rho = 0; end                   %相关系数

if nargin < 4, sig2 = 2;end                   %正态分布2方差
if nargin < 3, sig1 = 1;end                   %正态分布1方差

if nargin < 2, u2 = 2;end                     %正态分布2均值
if nargin < 1, u1 = 6;end                     %正态分布1均值

%global ava;
ava = [u1,u2];                                %高斯分布均值向量

cov_xy = rho*sig1*sig2;                       %协方差

%global sigma;
sigma = [sig1 cov_xy ;cov_xy sig2 ];          %协方差矩阵

%------数据显示网格范围------------%

scop1_l = u1-sqrt(sig1)-2;
scop1_r = u1+sqrt(sig1)+2;
scop2_l = u2-sqrt(sig2)-2;
scop2_r = u2+sqrt(sig2)+2;
[X,Y] = meshgrid(scop1_l:0.2:scop1_r,scop2_l:0.2:scop2_r);
xy = [X(:) Y(:)];
p = mvnpdf(xy,ava,sigma);                     %联合概率密度
P = reshape(p,size(X));

mesh(X,Y,P);                                  %3-D概率密度图形
name1 = ['二维正态分布 N(',num2str(u1),',',num2str(u2)];
name2 = [',',num2str(sig1),',',num2str(sig2),',',num2str(rho),')'];
name = [name1,name2];
title(name);
end