NEED EXPLANATION FOR A MATLAB AUDIO COMPRESSION URGENTLY PLEASE. I WOULD BE VERY GRATFUL

%%%%%%%%%%%%%TASK 1%%%%%%%%%%%%%%%%
[x,fs]=audioread('sample.wav');
N=length(x);
vlcplayer=audioplayer(x,fs);
vlcplayer.play

%%%%%%%%%%%% task 2%%%%%%%%%%%
t=fft(x,N);
X=fftshift(t)
f=-fs/2:fs/N:(fs/2-fs/N);
figure(1)
plot(f,abs(X))
title('original audio signal')


%%%%%%%%%%%%%% TASK 2 AND 4%%%%%%%%%
Xr=zeros(1,N);
Xr((N/4)+1:(3*N/4))= X((N/4)+1:(3*N/4));   %%FORMULA
figure(2)
plot(f, abs((Xr)));
xr= real(ifft(fftshift(Xr))); %%reconstruction
audiowrite('50% compressed.wav',xr,fs);
title('50% compressed audio')
xlabel('freq(hq)');ylabel('magnitude');

%%change ratio to 60,70,80,90,95% compression ....just change FORMULA
%Xr((N*((60/100)/2))+1 : N*(1-(60/100)/2)) = X((N*((60/100)/2))+1 :  N*(1-(60/100)/2));
%Xr((N*((70/100)/2))+1 : N*(1-(60/100)/2)) = X((N*((60/100)/2))+1 :  N*(1-(60/100)/2));
%Xr((N*((80/100)/2))+1 : N*(1-(60/100)/2)) = X((N*((60/100)/2))+1 :  N*(1-(60/100)/2));
%Xr((N*((90/100)/2))+1 : N*(1-(60/100)/2)) = X((N*((60/100)/2))+1 :  N*(1-(60/100)/2));
%Xr((N*((95/100)/2))+1 : N*(1-(60/100)/2)) = X((N*((60/100)/2))+1 :  N*(1-(60/100)/2));

%%PLAYING THE OUTPUT
%audiowrite('60% compressed.wav',xr,fs);
%audiowrite('70% compressed.wav',xr,fs);
%audiowrite('80% compressed.wav',xr,fs);
%audiowrite('90% compressed.wav',xr,fs);
%audiowrite('95% compressed.wav',xr,fs);

[Walter Roberson]

https://www.mathworks.com/matlabcentral/answers/29922-why-your-question-is-not-urgent-or-an-emergency

[Maitry Dumaswala]

Please help me with this error?

clc;
clear all;
close all;
a=imread('C:\Users\hp\Pictures\Saved Pictures\boat.jpg'); % to read the grayscale image 
b = imnoise(a,'salt & pepper'); % to create the noisy image 
figure
subplot (1, 2, 1) 
imshow(a) 
title ('Original Image') 
subplot (1, 2, 2) 
imshow(b) 
title ('Noisy Image') 

[Walter Roberson]

File C:\Users\Pictures\Saved Pictures\boat.jpg does not exist. You should look through your file system to determine whether you made a mistake in the directory name or the file name.

[Rakesh]

could you please forward the information about this code in pdf format like its abstract,introduction,references,theoritical analysis,etc.....like these

[Walter Roberson]

Rakesh, you are assuming that someone has already written those out for this code, which is not at all certain. What you are asking for sounds quite a bit like what a student would be expected to write for an assignment; if someone had already written such a thing up, that would disqualify a second student from using the results in their own assignment because of academic integrity concerns.

Answer 1

[Walter Roberson]

It is a high-pass filter. When you fft() the coefficients for the low frequencies come first, and increasing frequencies after that in the result until you get to the middle and then they reduce again. Middle is highest frequency, next to that on both sides is second highest frequency, and so on.

When you copy the middle of the results out to zeroes around it, then you are zeroing the lower frequencies, which is a high pass filter.

[HARSH MANDALAYWALA]

sorry to trouble you, but can you please explain it in little bit more detail on how it is related with audio compression?

[Walter Roberson]

You can store just the extracted parts, together with the sampling frequency, and the number of original samples, wihtout storing the zeros. Later you can use the information about the number of original samples and the length of what you did store in order to create a vector of zeros that you save the values into; with that and the sampling frequency, you can then play the (filtered) sound.

By the way, if you were trying to compress speech, then you would be better off storing from the left and right edges instead of from the middle. Most of the key frequencies for speech are lower frequencies instead of higher frequencies. Higher frequencies give "tone", but lower frequencies are used for words.

[HARSH MANDALAYWALA]

could you please provide the code by the using the method you just explained, it would be easy to understand that way. Thank you in advance.

[Walter Roberson]

For low-pass instead of high pass:

p=floor(N/4);
L=length(X);
r = [1:p+1,L-p+1:L];
Xr(r) = X(r);

[divya reddy]

@walter can you please explain how that "formula" thing in the code works?

[Walter Roberson]

When you fft() the coefficients for the low frequencies come first, and increasing frequencies after that in the result until you get to the middle and then they reduce again. Middle is highest frequency, next to that on both sides is second highest frequency, and so on.

Formulas such as

X((N*((60/100)/2))+1 :  N*(1-(60/100)/2))

are starting at the beginning and skipping through to (60/100)/2 of the way through the ascending frequency information, skipping the first 60/100 of the ascending frequency information. The 1-(60/100)/2 is going 60/100 of the way back from the end, skipping the last 60/100 of the descending frequency information. You are selecting the last 40 parts out of 100 of ascending frequency information, and the first 40 parts out of 100 of the descending frequency information -- 60% compression if you only store the middle parts.

[divya reddy]

X=fftshift(t)

xr= real(ifft(fftshift(Xr))); %%reconstruction

@walter thank you so so much!!!! :)

last doubt,

can you explain what role these two lines play in the code^^^^^

[Walter Roberson]

https://www.mathworks.com/help/matlab/ref/fftshift.html

Ah, I missed the fftshift() before. The effect is to make the algorithm into a low-pass filter instead of a high-pass filter.