加权平均、EMD、小波等方法去噪效果对比

加权平均、EMD、小波等方法去噪效果对比

代码

整体代码如下

%%
clear all;
clc;

load('data_filter120Hz.mat'); %可自己生成随机噪声
fs=1000;%采样频率是1000Hz
%%
%生成正弦波信号
t=linspace(0, length(data)/fs-1/fs, length(data));
y1 =15*sin(2*pi* 2.8 *t);%生成频率为2.8Hz，幅值为15的正弦波
y2 =10*sin(2*pi* 10.5 *t);%生成频率为10.5Hz，幅值为10的正弦波
y3 =3*sin(2*pi* 27 *t);%生成频率为27Hz，幅值为3的正弦波
y4 =0.5*sin(2*pi* 43 *t);%生成频率为43Hz，幅值为0.5的正弦波

y_Sin =y1+y2+y3+y4;
% y_Sin =y1+y2+y3;

y = y_Sin'+data;
%%
saveTime =1;
signal = y(1:saveTime*fs);
dlmwrite('MultiSinWaveWithNoise_1s.txt',signal,'delimiter',' ');

saveTime =10;
signal = y(1:saveTime*fs);
dlmwrite('MultiSinWaveWithNoise_10s.txt',signal,'delimiter',' ');

%%
figure;
subplot(3,1,1);
plot(data);
title('noise');
subplot(3,1,2);
plot(y_Sin);
title('sin wave');
subplot(3,1,3);
plot(y);
title('sin wave + noise');


figure;
plot(y_Sin);
hold on;
plot(y);
legend({'raw','with noise'});

%%
p=0.9;
preTemp =0;
for i=1:length(y)
    y_winAve(i) = preTemp*(1-p)+ y(i)*p;
    preTemp = y_winAve(i);    
end

error_noise = sum(abs(data));
error_winave = sum(abs(y_winAve -y_Sin));

%%
figure;
plot(y_Sin);
hold on;
plot(y_winAve);
legend({'raw','win ave'});

figure;
plot(y);
hold on;
plot(y_winAve);
legend({'raw+noise','win ave'});


%% emd method
emd_num = 2;

imf = emd(y);
y_emd =sum(imf(:,emd_num:end),2);
figure;
plot(y);
hold on;
plot(y_emd);
title('emd denoise');

error_emd = sum(abs(y_emd' -y_Sin));

%% wpdencmp
wpden_num =3;

[thr,sorh,deepapp,crit]=ddencmp('den','wp',y);
[y_wpden,~,~,~]=wpdencmp(y,sorh,wpden_num,'sym6',crit,thr,deepapp);
figure;
plot(y);
hold on;
plot(y_wpden);
title('wpdencmp');

error_wpden = sum(abs(y_wpden' -y_Sin));


%% winave 2nd method
p2 =0.85;
winLen =10;
preTemp =0;
for i=1:length(y)
    if(i<length(y)-winLen)
        if(winLen>=i)
            preTemp =mean(y(1:i+winLen));
        else
            preTemp =mean(y(i-winLen:i+winLen));
        end
    else
        preTemp =mean(y(i-winLen:length(y)));
    end
    
    y_winAve02(i) = preTemp*(1-p2)+ y(i)*p2;
    preTemp = y_winAve02(i); 
end

error_winave02 = sum(abs(y_winAve02 -y_Sin));

figure;
plot(y);
hold on;
plot(y_winAve02);
legend({'raw+noise','win ave 02'});

%%
close all;

效果

从结果上去看，上述参数中，去噪效果：
小波 >EMD >加权平均 >移动平均（具体设置看上方代码）

Matlab转c++

emd和小波去噪的C++代码效果和matlab自带的效果不太一致（可能是我设置的问题），但都能达到去噪的效果，此时emd效果最好，小波的效果需要调整软阈值的值来优化（0.5->1）；

EMD代码
小波代码