有监督的蚁群算法图像分割

clear all; close all; clc; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%产生作为输出缓噪声的伪噪声序列%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % temp=(randn(1,8192)>0); % PN=zeros(1,16384); % PN(1:8192)=temp; % PN(16384:-1:8193)=-temp; % w_k=zeros(1,16384); % w_k(1:3715)=1; % w_k(end-3715:end)=1; % FFT_value=w_k.*exp(j*PN*pi); % FFT_temp(1)=0; % FFT_temp(2:16385)=FFT_value; % IFFT_temp=ifft(FFT_temp,16385); % IFFT_value=IFFT_temp(2:16385); % FINAL_value=fftshift(real(IFFT_value)); % FINAL_value=resample(FINAL_value,2,11); % PN_value=FINAL_value; % PN_num=1; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%产生一个回声路径的冲激响应%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% N = 240; %240个点 t = [0 : N-1]'; %0：1：239 tau = N./4; %tau=60 envelope = exp(-(t./tau));%随即信号的幅度权值 w_k = randn(N,1).*envelope; w_k = [w_k(end-19 : end);w_k(1 : end-20)]; %把后20个点放到前面 w_k = w_k./sqrt(N);%再次乘上一个固定的增益 figure; plot(w_k);title('回声路径的冲击响应') %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%输入两段语音信号作为远端和近端信号%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % array1=wavread('quiet',100000);%近端语音 % array2=wavread('wang2',100000);%远端语音 %array1=wavread('wang1',300000);%1 2 ... array1=wavread('最初的梦想CD音质.wav',500000);%1 2 ... 采样频率 44100Hz 近端信号 % array2=wavread('wang2',300000); % array1(1:100000)=wavread('quiet',100000); % array2(end-69999:end)=wavread('quiet',70000); % array1=wavread('quiet',300000); %array2=wavread('wang2',300000); %10 11 .... array2=wavread('现在就想见的人CD音质.wav',500000); %10 11 .... 远端 信号 Rin2=array2(:,2)'; array1=array1(:,1)'; array2=array2(:,1)'; %wavplay(array1,22050); %wavplay(array2,22050); M=length(array2); %序列总长度 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%通过远端语音信号与回声路径冲激响应卷积生成回声信号%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% N=length(w_k);%滤波器阶数 240个点 %pecho=zeros(1,N); pecho=conv(array2,w_k); %%卷积处理后 长度变为M+N-1 pecho=pecho(1:M);%模拟的回声信号 取出与原始语音信号相同的点数 %array1= 0.05*mean(abs(pecho))*randn(1,M); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%读入N个数据，设置初值%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Rin1=array2; %远端信号 作为滤波器的输入信号 % Sinn=pecho; Sinn=2*pecho+array1;%近端信号——近端说话人的语音信号与回声叠加构成 作为期望信号 Sinn=Sinn/max(Sinn); % wavplay(pecho,8000); % wavplay(Sinn,8000); x1=Rin1(1:N);%？？？ w1=zeros(1,N);%%%滤波器系数的初始化 w2=zeros(1,N); a=0.0001; u=0.01;%%步长 energ_window=1/32; %短窗功率估计 near_energ=0; far_energ(1)=0; for i=2:N-1 far_energ(i)=(1-energ_window)*abs(far_energ(i-1))+energ_window*abs(x1(i-1));%初始化远端信号的能量 end; Dtemp=x1(1:N-1)*x1(1:N-1)';%滤波器输入信号的能量 sout=zeros(1,M); ff=0; dd=0; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%NLMS算法的自适应滤波%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% for i=N:M; x1(N)=Rin1(i); x2(N)=Rin2(i); D=Dtemp+x1(N).^2; %%D 跟Dtemp 用来干嘛？？ far_energ(N)=(1-energ_window)*abs(far_energ(N-1))+energ_window*abs(x1(N-1));%远端信号能量 near_energ=(1-energ_window)*abs(near_energ)+energ_window*abs(Sinn(i-1));%近端信号能量 y1=dot(w1,x1); y2=dot(w2,x2); en=Sinn(i)-y1-y2; if near_energ>1/2*max(far_energ); %%%%具有对讲保护的功能 sout(i)=en; dd=dd+1; % sout(i)=nlp_new(en); ;%远端模式，启动NLP else w1=w1+u*en*x1/(a+D); % D Dtemp 是滤波器输入信号的能量！！！ w2=w2+u*en*x2/(a+D); sout(i)=en; % sout(i)=nlp_new(en); ;%远端模式，启动NLP ff=ff+1; end; % mis(i)=norm(w1'-w_k)/norm(w_k); %%???? %%%%%%%%%更新滤波器输入信号的能量值%%%%%%%%%% Dtemp=D-x1(1).^2; x1(1:N-1)=x1(2:N); x2(1:N-1)=x2(2:N); far_energ(1:N-1)=far_energ(2:N);%%%更新远端信号的能量 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% end; % figure; % plot(array1); % hold on; % plot(pecho,'r'); % figure; % plot(Sinn); figure; plot(Rin1);title('自适应滤波器的输入信号A路') figure; plot(Rin2);title('自适应滤波器的输入信号B路') figure; plot(array1);title('近端信号') figure; plot(Sinn); legend('near end signal');hold on; %画出期望信号时域波形 figure;plot(sout/max(sout),'r');legend('error signal');%画出输出信号的时域波形 wavwrite(Sinn,44100,16,'Sinn');%%自适应滤波器的期望信号 % wavwrite(Rin,44100,16,'Rin');%%自适应滤波器输入信号 wavwrite(sout,44100,16,'sout');%%自适应滤波器的输出信号 figure;subplot(4,1,1);plot(sout);title('输出的误差信号'); subplot(4,1,2);plot(array1);title('近端信号') subplot(4,1,3);plot(Sinn);title('近端信号与回波的混合') subplot(4,1,4);plot(array1);title('回波衰减比率') % wavwrite(Sinn,8000,16,'Sinn'); % wavwrite(Rin,8000,16,'Rin'); % wavwrite(sout,8000,16,'sout'); %wavplay(sout,22050); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%性能测试%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Hd2 = dfilt.dffir(ones(1,1000));%%%%设计的滤波器？？？ erle = filter(Hd2,(sout-array1(1:length(sout))).^2)./ ... (filter(Hd2,Sinn(1:length(sout)).^2)); erledB = -10*log10(erle); plot(erledB);title('回波的衰减比率') % P_Sinn=abs(Sinn).^2; % P_en=abs(sout).^2; % Ave_num=100;%把原始语音信号分帧 每5600个点为一帧 % P_num=fix(M/Ave_num); % ERLE_temp=P_Sinn./P_en; %期望信号的能量比输出信号的能量 % for j=1:P_num; % Mark=(j-1)*Ave_num; % ERLE_win=ERLE_temp(Mark+1:Mark+Ave_num); % ERLE(j)=sum(ERLE_win)/Ave_num; % end; % figure; % plot(ERLE); % ERLE=10*log10(ERLE);%取自然对数 % figure; % plot(ERLE); % figure; % plot(10*log10(mis)); % plot misalignment curve % ylabel('mis');xlabel('samples'); % title('Misalignment of NLMS algorithm');