A list,

Speech enhancement refers to the technology of extracting useful speech signals from the noise background to suppress and reduce the noise interference when the speech signal is disturbed or even submerged by various noises. In a word, extract the original speech as pure as possible from the noisy speech. The actual speech interference can be divided into the following categories: (1) periodic noise, such as electrical interference, interference caused by engine rotating part, such as some discrete narrow frequency peaks; ② Impulse noise, such as some electric spark, discharge noise interference; ③ Wideband noise, which refers to gaussian noise or white noise, which is characterized by wide frequency band, almost covering the whole speech band; (4) voice interference, such as picking up other people’s speech in the microphone, or transmission encountered crosstalk caused by speech. To deal with the different types of noise mentioned above, the enhancement techniques are also different. One of the main goals of speech enhancement is to extract the original speech as pure as possible from the noisy speech signal. However, since the interference is usually random, it is almost impossible to extract completely pure speech from noisy speech. In this case, the purpose of voice enhancement is mainly two: one is to improve voice quality, eliminate background noise, so that the listener is happy to accept, do not feel tired, which is a subjective measure; Second, it’s an objective measure. These two purposes are often incompatible. At present, there are some speech enhancement methods for low SNR noisy speech, which can significantly reduce the background noise and improve the speech quality, but can not improve the speech intelligibility, even slightly decreased. Speech enhancement is not only related to the theory of speech signal digital processing, but also involves human auditory perception and phonetics. In addition, noise comes from many sources and their characteristics vary from one application to another. Therefore, different speech enhancement strategies must be adopted for different noises. Some speech enhancement algorithms have been proved to be effective in practical applications. They can be roughly divided into four categories: noise elimination method, harmonic enhancement method, speech reconstruction method based on parameter estimation and speech enhancement algorithm based on short-time spectrum estimation.

Ii. Source code

%% P2_2_1 speech signal sampling, draw the time domain and frequency domain graph [Y, FS] = Audioread ('myname.wav'); % y is the sampled data; Fs is the sampling frequency sound(y,fs); % Play language signal myname N = length(y); % Signal length t = (0:N- 1)/fs; % Time domain figure(1); subplot(211); plot(t,y); title('Audio Time domain map'); % Audio signal time domain map xlabel('Time'); ylabel('Amplitude'); Y = fft(y,N); % FFT Analysis spectrum df = fs/length(Y); % Calculate the spectral line interval f =0:df:(fs/2-df); % spectrum range, cut the first half (sampling frequency is higher than the maximum frequency2Yf = times)abs(Y); % amplitude response Yf = Yf(1:length(Yf)/2); % Since the amplitude response is an even function, cut half % to draw the figure(1);
subplot(212); axis([0,fs/5.0.5000]); plot(f,Yf); title('Audio spectrum map'); %% P2_2_2 time domain and frequency domain graph y = y(:,1); y = y';
nt = randn(1,length(y)); % generate Gaussian noise % design BPF get4KHZ and5KHZ noise FP1 =4000; fp2 =5000; % BPF indicator WP1 =2*pi*fp1/fs; WP2 =2*pi*fp2/fs; % Convert analog index to digital index Wn1 = [WP1 WP2]; b = fir1(34,wn1,'bandpass');
yn = filter(b,1,nt); % noise signal s = y+yn; % mix band limited noise with original audio %sound(s,fs); S = fft(s,N); Sf =abs(S); % amplitude DF = fs/length(S); % Calculate the spectral line interval f =0:DF:(fs/2-DF); % spectrum range, cut the first half (sampling frequency is higher than the maximum frequency2Times) Sf = Sf(1:length(Sf)/2); % Since the amplitude response is an even function, cut half % to draw the figure(2);
subplot(211); plot(t,s); title('Audio time domain map with noise added');
xlabel('Time'); ylabel('Amplitude'); grid on; subplot(212); axis([0,fs/5.0.10000]); plot(f,Sf); title('Audio spectrum with noise added');
xlabel('Frequency'); ylabel('Amplitude'); grid on; %% P2_2_3 design filter, and filter % design band stop filter to filter noise FP1 =3500; FS1 =3600; FP2 =5600; FS2 =5500;
rp2 = 1; rs2 =40; FP = [FP1 FP2]; FS = [FS1 FS2]; wp =2*pi*FP/fs; % converts the analog metric to the digital metric ws =2*pi*FS/fs; % Chebyshev IIR filter [N,wn] = Cheb1ORD (WP, WS, RP2,rs2); % the minimum order of filter is n, wn is system frequency band [bz,az] = cheby1(n,rp2, WP,'stop'); sound(yfilt,fs); % Play the filtered voice signal Yfil =abs(fft(yfilt,N));
Yf = Yfil(1:length(Yfil)/2);
Copy the code

3. Operation results





Fourth, note

Version: 2014a complete code or write plus 1564658423