目的:
1.通过窗函数法设计FIR滤波器
2.使用窗函数法设计FIR滤波器,了解窗函数的形式和长度对滤波器性能的影响。
实现环境:
Matlab
理想低通滤波器设计:
function my_output=ideallp(wc,N)
%Ideal Lowpass filter computation
%------------------------------------
%[hd]=ideal_lp(wc,M)
% hd=ideal impulse response between 0 to M-1
% wc=cutoff frequency in radians
% M=length of the ideal filter
%
alpha=(N-1)/2;
n=0:1:(N-1);
m=n-alpha+eps;
my_output=sin(wc*m)./(pi*m);
end
wp=0.2*pi;
ws=0.3*pi;
deltaw=ws-wp;
N=ceil(1.8*pi/deltaw);
wc=(wp+ws)/2;
hd=ideallp(wc,N);
wd1=boxcar(N)';
h1=hd.*wd1;
[H1,w]=freqz(h1,1);
plot(w/pi,20*log10(abs(H1)));
title('矩形窗');
得到频率响应曲线如下:
wp=0.2*pi;
ws=0.5*pi;
As=50;
deltaf=(ws-wp)/(2*pi);
N=ceil((As-7.95)/(14.36*deltaf))+1; %
beta=0.1102*(As-8.7);
wdkai=kaiser(N,beta)';
wc=(ws+wp)/2;
hd=ideallp(wc,N);
h=hd.*wdkai;
[H,w]=freqz(h,1);
plot(w/pi,20*log10(abs(H)));
title('凯泽窗');
得到频率响应图像如下:
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.1*pi/deltaw);
wdtri=triang(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdtri;
[H,w]=freqz(h,1);
plot(w/pi,20*log10(abs(H)));
title('Triang');
得到频率响应图像如下:
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.6*pi/deltaw);
wdhamm=hamming(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdhamm;
[H,w]=freqz(h,1);
plot(w/pi,20*log10(abs(H)));
title('Hamming');
得到频率响应图像如下:
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.1*pi/deltaw);
wdblack=blackman(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdblack;
[H,w]=freqz(h,1);
plot(w/pi,20*log10(abs(H)));
title('Blackman');
得到频率响应图像如下:
一、 主瓣的宽度由N决定,旁瓣的衰减由窗函数的形状决定。
二、 Kaiser的不同在于可以同时调整主瓣宽度和旁瓣衰减,这是其他窗函数不具备的。
三、 实际滤波的应用
设置初始信号为10Hz的余弦与40Hz余弦叠加而成的信号,分别采用以上滤波方法滤波。
%初始信号
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
subplot(2,1,1);
plot(t,y);
title('initial signal');
yfft=fftshift(fft(y));
w=-50:0.5:50;
subplot(2,1,2);
plot(w,abs(yfft));
axis([0 50 0 100]);
title('frequence');
%滤波器内容
wp=0.2*pi;
ws=0.3*pi;
deltaw=ws-wp;
N=ceil(1.8*pi/deltaw);
wc=(wp+ws)/2;
hd=ideallp(wc,N);
wd1=boxcar(N)';
h1=hd.*wd1;
final=fftfilt(h1,y,200);
%滤波之后的信号
figure;
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
运行结果如下:
初始信号如下
处理后信号如下:
可见40Hz分频成分明显降低,而10Hz分频保留较好。
同理,更改代码中的滤波器部分分别得到下面几个滤波结果对比
2. Kaiser
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
w=-50:0.5:50;
wp=0.2*pi;
ws=0.5*pi;
As=50; %
deltaf=(ws-wp)/(2*pi);
N=ceil((As-7.95)/(14.36*deltaf))+1; %
beta=0.1102*(As-8.7);
wdkai=kaiser(N,beta)';
wc=(ws+wp)/2;
hd=ideallp(wc,N);
h=hd.*wdkai;
final=fftfilt(h1,y,200);
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
w=-50:0.5:50;
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.1*pi/deltaw);
wdtri=triang(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdtri;
final=fftfilt(h1,y,200);
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
w=-50:0.5:50;
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.2*pi/deltaw);
wdhann=hanning(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdhann;
final=fftfilt(h1,y,200);
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
w=-50:0.5:50;
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.6*pi/deltaw);
wdhamm=hamming(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdhamm;
final=fftfilt(h1,y,200);
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
t=0:0.01:2;
y=cos(2*pi*10*t)+cos(2*pi*40*t);
w=-50:0.5:50;
wp=0.2*pi; %
ws=0.3*pi;
As=50;
deltaw=ws-wp;
N=ceil(6.1*pi/deltaw);
wdblack=blackman(N)';
wc=(wp+ws)/2;
hd=ideallp(wc,N);
h=hd.*wdblack;
final=fftfilt(h1,y,200);
subplot(2,1,1);
plot(final);
title('processed signal');
ynfft=fftshift(fft(final));
subplot(2,1,2);
plot(w,abs(ynfft));
axis([0 50 0 100]);
title('frequence filted');
不同滤波方式衰减过程不同,但由于此次信号只有两个分频,体现的不太明显。