我的任务是要求将音频信号与噪声信号相结合，并使用数字滤波器使原始音频信号不失真。
问题是我无法获得原始音频，因为我无法估计频域中的噪声内容。我用了10阶的巴特沃斯低通滤波器的试验，试图猜测截止频率，但总是产生的文件仍然很嘈杂。
任何帮助？
以下是我的步骤：
%% Load the original audio file to matlab Workspace
test_audio ='J:\test_read.wav';
%% Load the noise audio file to Matlab workspace
noise ='J:\crowd_noise.wav';
%% read the original audio file for a finite length of time
[y,Fs] = audioread(test_audio,[1 262144]);
%%read the noise audio file for a finite length of time
[x,fs]=audioread(noise,[1 262144]);
[p,q]=rat(Fs/fs);
%% resample the noise file with the original file's sampling rate
m=resample(x,p,q);
m=m(:,1); %% take only one channel
Y=y(1:length(m),1); %% make the 2 streams of equal lengths
u=m+Y; %% the resultant noisy signal
%% DFT of the resultant signal
N=length(u);
df = Fs / N;
w = (-(N/2):(N/2)-1).*df;
U= fft(u,N) / N; %For normalizing, but not needed for our analysis
U = fftshift(U);
figure;
plot(w,abs(U);
spectrogram(u,256,[],[],Fs);colorbar
%% DFT of the original signal
df = Fs / N;
w = (-(N/2):(N/2)-1).*df;
k= fft(y,N) / N; %For normalizing, but not needed for our analysis
K= fftshift(k);
figure;
plot(w,abs(K);
spectrogram(y,256,[],[],Fs);colorbar
[psdy,Fyy] = periodogram(y,rectwin(length(y)),length(y),Fs);
plot(Fyy,10*log10(psdy));
grid on;
以下链接包含FFT图表和音频文件：
提前致谢。