MCRA2+过减法降噪

最新推荐文章于 2024-07-24 18:01:02 发布

NosONE

最新推荐文章于 2024-07-24 18:01:02 发布

阅读量129

点赞数

分类专栏：笔记文章标签：实时音视频算法

本文链接：https://blog.csdn.net/amateurSU/article/details/132408201

版权

笔记专栏收录该内容

7 篇文章 0 订阅

订阅专栏

MCRA2

总体步骤：
在这里插入图片描述

跟踪最小值采用的“连续频谱最小值跟踪算法
在这里插入图片描述

过减技术的谱减法

在这里插入图片描述

Matlab代码

% mcra2 + Spectral Subtraction
% 2023.8.14  Author: Sfj
clc;clear;close all;

% %%%%%%%%%%%%%%%%- Read input-file  %%%%%%%%%%%%%%
inputname= 'test_wav';
path= 'data/';
read_pcm= 0;
outpath= ['data/out0/'];
mkdir(outpath);

Channel= 1;

if (read_pcm== 1)
    fileId= fopen([path,inputname,'.pcm'],'r');
    x= fread(fileId, inf,'int16');
    fclose(fileId);
    for i= 1:Channel;
        x_in(:,i)= x(i:Channel:end);
    end
else
    [x_in,fs]= audioread([path,inputname,'.wav'],'native');
    x_in= x_in(:,Channel);
end
x= double(int16(x_in));

frame_size= floor(16 * fs / 1000) % 16ms,256
FFT_size= 2 * frame_size;
FFT_bins= frame_size + 1;
x_old= zeros(frame_size,Channel);
%win= hanning(FFT_size);
win = (sin(pi*(0.5:1:FFT_size-0.5)./FFT_size)).';

w=2;
b=hanning(2*w+1);
b=b/sum(b);
track_min = 0; % 选择搜索最小值的方法，滑动窗口找最小值
if track_min == 1
  Nwin= 8;
  Vwin= 15;
  l_mod_lswitch=0;
  Smin_sw = zeros(FFT_bins,Nwin);s
end

%  Initination
freq_res= fs / FFT_size;
k_1khz= floor(1000 / freq_res);
k_3khz= floor(3000 / freq_res);
delay_val= [2*ones(k_1khz,Channel);2*ones(k_3khz-k_1khz,Channel);5*ones(FFT_bins-k_3khz,Channel)];
alpha_s_mc2= 0.8;
gamma_mc2= 0.998;
beta_mc2= 0.96;
alpha_p_mc2= 0.2;
alpha_d_mc2= 0.95;
beta= 0.002;
img= sqrt(-1);

S_old= zeros(FFT_bins,Channel);
Smin_old= zeros(FFT_bins,Channel);
phat= zeros(FFT_bins,Channel);
lambda_dav= zeros(FFT_bins,Channel);
x_frame_old= zeros(frame_size,Channel);
function a=berouti(SNR)

if SNR>=-5.0 & SNR<=20
   a=4-SNR*3/20;
else

  if SNR<-5.0
   a=5;
  end

  if SNR>20
    a=1;
  end

end
endfunction

h = waitbar(0,'Running...');
set(h,'name','mcra2_spsub - 0%');
nb_frames= fix((size(x,1) - frame_size) / frame_size);
for frame_idx= 1:nb_frames
    pos= frame_size * (frame_idx-1) + 1;
    x_new= x(pos:pos+frame_size-1,:);
    x_frame= [x_old;x_new];
    x_old= x_new;

    temp= fft((x_frame.*win));
    Y= temp(1:FFT_bins);
    Y_mag= abs(Y);
    Ya2= Y_mag.*Y_mag;
    Yf= conv(b,Ya2);  % _mcra2_out3 add
    Yf= Yf(w+1:FFT_bins+w);
% mcra2
%   Continuous Spectral Minimum Tracking
    S= alpha_s_mc2 * S_old + (1-alpha_s_mc2) * Yf;
    Sminc= S;
    Idx= find(Smin_old < S);
    Sminc(Idx)=  gamma_mc2.*Smin_old(Idx) + (1-gamma_mc2)/(1-beta_mc2).*(S(Idx)-beta_mc2.*S_old(Idx)); %连续谱找最小值
    S_old= S;
    Smin_old= Sminc;

    if(track_min==1)
      if frame_idx < Vwin
        Smin = S;
        SMact = S;
      endif
      Smin = min(Smin,S);
      SMact = min(SMact,S);
    end
    Smin_final = Sminc; % 最终参与计算的最小值
%  Calculate the probability of speech presence
    p= zeros(FFT_bins,Channel);
    Idx2= find(S>(Smin_final.*delay_val));
    p(Idx2)= 1.0;
    phat= alpha_p_mc2.*phat + (1-alpha_p_mc2).*p;
    alpha_dt= alpha_d_mc2+(1-alpha_d_mc2)*phat;
    lambda_dav= alpha_dt.*lambda_dav+(1-alpha_dt).*Ya2; % Estimated Noise Magnitude Spectrum
    lambda_d=1.4685*lambda_dav; % _mcra2_out4 add
    if(track_min==1)
      if frame_idx < Vwin
        l_mod_lswitch = Vwin-1;
      end
        l_mod_lswitch=l_mod_lswitch+1;
        if l_mod_lswitch==Vwin
          Smin_sw= [Smin_sw(:,2:end),SMact];
          Smin = min(Smin_sw,[],2);
          SMact = S;
        endif
    end
% Spectral Subtraction
    noise_mag= sqrt(lambda_d);
    phase= angle(Y);%atan2(imag(Y),real(Y)); % 弧度
    SNRseg= 10*log10(norm(Y_mag,2)^2/norm(noise_mag,2)^2);
    alpha_sub= berouti(SNRseg);

    Xpower= Ya2-alpha_sub.*lambda_d;
    z= find(Xpower < beta*lambda_d);
    Xpower(z)= beta*lambda_d(z);

    % Xpower = max(Ya2-alpha_sub.*lambda_d, beta*lambda_d);
    X_mag= sqrt(Xpower);

    X= X_mag.*(cos(phase)+ img*sin(phase));
% IFFT and OLA
    X(FFT_bins+1:FFT_size)=conj(X(FFT_bins-1:-1:2));
    x_frame_out= win.*real(ifft(X));
    x_out_frame=  x_frame_out(1:frame_size,:)+x_frame_old;
    x_frame_old= x_frame_out(frame_size+1:end,:);

    NS_OUT(pos:pos+frame_size-1)= x_out_frame;

    waitbar(frame_idx / nb_frames,h);
    set(h,'name',[sprintf('%2.1f',frame_idx / nb_frames*100)]);
end

close(h);
audiowrite([outpath,inputname,'_mcra2_out5','.wav'],NS_OUT/32767,fs);