matlab 小波滤波器,matlab小波滤波器使用

研究db小波，用分解和重构滤波器和上下采样函数实现多分辨分析，代码如下：

%%% 小波分解与重构

clear;close all;

load noissin;Sig=noissin;

%% 滤波器分解

[Lo_D,Hi_D,Lo_R,Hi_R] = wfilters('db3');

[c,l] = wavedec(Sig,2,Lo_D,Hi_D);

[cd1,cd2]=detcoef(c,l,[1,2]);

ca2 = appcoef(c,l,'db3',2);

% 重构

X=waverec(c,l,Lo_R,Hi_R);

%% 换种方式分解

[C,L] = wavedec(Sig,2,'db3');

[Cd1,Cd2]=detcoef(c,l,[1,2]);

Ca2 = appcoef(c,l,'db3',2);

Ca1=appcoef(c,l,'db3',1);

%% 手动卷积分解

det1=conv(Hi_D,Sig);

det1=dyaddown(det1);

app1=conv(Lo_D,Sig);

app1=dyaddown(app1);

det2=conv(Hi_D,app1);

det2=dyaddown(det2);

app2=conv(Lo_D,app1);

app2=dyaddown(app2);

% 重构

App1=conv(Lo_R,dyadup(app2))+conv(Hi_R,dyadup(det2));

App1=App1(6:end-5); % 前后各去掉5个值

Xs=conv(Lo_R,dyadup(App1))+conv(Hi_R,dyadup(det1));

% figure(1);subplot(211);plot(Cd1);title('cd1')

% subplot(212);plot(det1);title('det1')

% figure(2);subplot(211);plot(Ca1);title('ca1')

% subplot(212);plot(app1);title('app1')

卷积后发现结果的头尾几个元素不对，可能是延拓的方式有区别，所以手动重构时去掉了5个，是滤波器长度-1，最后结果除头尾不对，其它一致。