How can convert vector aa size 1x262144 to matrix 256x256

hi every one i'm new her i need help

size of aa 1x262144 i want convert to matrix 256x256

[honey]

how can we calculate the frequency of specific signal in matlab

[Walter Roberson]

In order to calculate the frequency of a signal, you need to know either the time between samples or else the sampling rate (which is just the inverse of the time between samples.) You can then proceed to fft(), and find the peak of that (you probably want to ignore the first bin, and you probably want to look at abs()). Then convert the bin location into frequency; the first example in the fft() documentation shows you how to do that.

You might also want to consider using spectrogram()

[honey]

[y,fs]=audioread("filename");
disp(fs)

but the above didn't print the sampeled values

if there are anotherways for this plzz describe

along with these why we choose 1*262144 as size. is there any reason for it

[Walter Roberson]

In your posted code, fs would be the sampling frequency. The sampled values would be in y. You would proceed to fft(y) and then analyze the results along the lines I mentioned.

Note: your y might have multiple channels, which will show up as separate columns. Be careful, as in theory the different channels can have different peak frequencies.

Answer 1

[Isabella Osetinsky-Tzidaki]

 % you will not convert (262144 is 4 times 256*256 and not equal to it)
% but complete 4 matrices out of your "aa" vector 
% and you will get a "M" array:
L=256*256;
M=nan(256,256,4); % pre-allocate
for i=1:4
   v=aa(L*(i-1)+1:L*i);
   M(:,:,i)=reshape(v,256,256);
end

[honey]

[y,fs]=audioread('filename');
x1=y(1:262144);
x1=reshape(x1,1,262144);

can you please explain what is happened in above function

and its significance also

[Walter Roberson]

The first line reads the given audio file. The samples are stored in columns, one column per channel. The samples are stored in the variable named y. The sampling frequency is stored in the variable fs.

The next line extracts the first 262144 samples from y into a vector. If the file had more than that many samples in the left (or only) channel then it simply extracts that many from the first column. The output in x1 is a column.

The third line reshapes the column x1 into a row x1.

[honey]

thank u so much

can u please explain this pseudo random sequence

z

%% UNCORRELATED PSEUDO RANDOM SEQUENCE GENERATOR
 
G=Key;
%Generation of first m-sequence using generator polynomial [45]
sd1 =[0 0 0 0 1];             % Initial state of Shift register
PN1=[];                       % First m-sequence
for j=1:G
    PN1=[PN1 sd1(5)];
    if sd1(1)==sd1(4)
        temp1=0;
    else temp1=1;end
    sd1(1)=sd1(2);
    sd1(2)=sd1(3);
    sd1(3)=sd1(4);
    sd1(4)=sd1(5);
    sd1(5)=temp1;
end
sd2 =[0 0 0 0 1];             % Initial state of Shift register
PN2=[];                       % Second m-sequence
for j=1:G
    PN2=[PN2 sd2(5)];
    if sd2(1)==sd2(2)
        temp1=0;
    else temp1=1;
    end
    if sd2(4)==temp1
        temp2=0;
    else temp2=1;
    end
    if sd2(5)==temp2
        temp3=0;
    else temp3=1;
    end
    sd2(1)=sd2(2);
    sd2(2)=sd2(3);
    sd2(3)=sd2(4);
    sd2(4)=sd2(5);
    sd2(5)=temp3;
end

[Walter Roberson]

Something about generator polynomials. That is potentially related to https://www.mathworks.com/help/comm/ref/cyclpoly.html or https://www.mathworks.com/help/comm/ref/rsgenpoly.html or https://www.mathworks.com/help/comm/ref/bchgenpoly.html

But if I were to guess, I would say that it looks like it is generating a pseudo-random stream of bits to be XOR'd with the input stream. The reason for that could potentially be encryption, but it is more likely to have to do with increasing robustness to noise, because if you have an input signal that happens to be constant, you can run into detection glitches if the signal does not change state often enough.

I have seen generator polynomials used in PHY interfaces for gigabit ethernet -- again because forcing the signal to change state gets you better detection characteristics than a signal that might be constant.

[honey]

thank you

in gray scale and binary image which gives the better result and why

[Walter Roberson]

If you are still working on the task of transmitting an image through an audio stream, then the answer is "It does not matter whether the image is binary or grayscale or color". However, for the same number of rows and columns, color images contain more information than grayscale or binary, and grayscale contain more information than binary, so for the same number of rows and columns of the image, color images would need the longest audio stream, and grayscale would need a shorter audio stream (by a factor of 3), and binary would need 1/8th of the audio stream that grayscale would need.

[honey]

thank you

how HAAR wavelet transform works on image in matlab

[Walter Roberson]

https://www.mathworks.com/help/wavelet/examples/haar-transforms-for-time-series-data-and-images.html

[honey]

  [L,H,V,D] = dwt2(x,'haar','sym');
  [CL,CH,CV,CD] = dwt2(H,'haar','sym');
  [CA1,CA2,CA3,CA4] = dwt2(CH,'haar','sym');

can you please explain how this works

[Walter Roberson]

Roughly speaking:

It has been found that for many situations, roughly 80% of the variation is due to 20% of the causes.

For example suppose you were doing an fft() analysis of a sound, then it would not be uncommon to find that 80% of the signal magnitude could be accounted for by knowing the coefficients for only 20% of the locations -- that if you were to zero out 80% of the fft coefficients and ifft() that the result would be "80%" similiar to the original -- so like 5:1 compression.

Furthermore, it turns out that for a lot of situations, that of the remaining 20% of the detail, that 80% of it can be accounted for by 20% of the remaining coefficients. So if you were to analyze the 20% of the signal that remained after the first round of compression, you could compress that by about 5:1 too, leaving about 20% of the 20% unaccounted for. For example if you started with a signal of length 100, and the first round found 20 coefficients that accounted for 80% of the signal, and then you too the remaining 20% of the signal then another 16 coefficients would account for 80% of the 20%, so 20+16 coefficients would account for 1-0.8*0.8 = 96% of the signal. The same 80/20 rule can be applied again multiple times, each time accounting for about 4/5 more of the detail.

That is the sort of approach this code is taking: it does a dwt2 haar transform and extracts four bands of coefficients. It then takes one of the four bands, the Horizontal detail, and analyzes it, getting out four bands of coefficients of that. It then analyzes the details of the horizontal detail -- because you can proceed to zero out most of [CA1,CA2,CA3,CA4] and store that in reduced form, and later expand those to the original size and use them to reconstruct CH, and use that to reconstruct x. This would get you a more accurate reconstruction than if you were just to do a single dwt2 and zero out a bunch of coefficients to do compression.

[honey]

thank you

but

[L,H,V,D] = dwt2(x,'haar','sym');

what is the significance of symmetric extension, x is speech signal

and also why we convert matrix to cell in image processing

for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
newCA2=mat2cell(CA2,a,b);
d=zeros(64,64);
nCA2=mat2cell(d,a,b);

[Walter Roberson]

I do not know why sym and haar are used.

sym appears to refer to "half-point symmetry" and is described at https://books.google.ca/books?id=Z76N_Ab5pp8C&pg=PA273&lpg=PA273&dq=Symmetric+extension+(half+point)&source=bl&ots=qWVKcu3y8-&sig=ACfU3U2GiZqB5FES0OPGzJ07L-e1qkEk_Q&hl=en&sa=X&ved=2ahUKEwjEu9jD8aLnAhXSup4KHaacBoIQ6AEwAXoECAsQAQ#v=onepage&q=Symmetric%20extension%20(half%20point)&f=false

haar is commonly used with audio because apparently it has good time localization.

I do not know why that cell stuff is being done. The code is dividing a 64 x 64 numeric array into cells of 4 x 4, but I do not know why it would want to do that.

[honey]

thank you

what is the significance after converting from matrix to cell in general in image processing

[Walter Roberson]

The main significance is that many people do not know how to write code for processing blocks of an image at a time, so they think they have to divide the image up into cell arrays in order to process it.

The second main significance is that sometimes the code needed to write code for processing blocks of an image can end up obscuring the algorithm being used, so sometimes it is clearer to convert into blocks and process the blocks using cellfun() instead of writing loops. Notice that I said "clearer", not "more efficient" or "better".

[honey]

okk thank you

[honey]

%% WATERMARK BIT ENCODING ALGORITHM IN SELECTED/MID-BAND COEFFICIENTS OF DCT TRANSFORMED BLOCKS ONLY
%% FOR BLOCK CA2
 
for p=1:16
    for q=1:16
        if w1(p,q)==0
            nCA2{p,q}(1,3)=nCA2{p,q}(1,3)+alpha*PN1(1);
            nCA2{p,q}(1,4)=nCA2{p,q}(1,4)+alpha*PN1(2);
            nCA2{p,q}(2,2)=nCA2{p,q}(2,2)+alpha*PN1(3);
            nCA2{p,q}(2,3)=nCA2{p,q}(2,3)+alpha*PN1(4);
            nCA2{p,q}(3,1)=nCA2{p,q}(3,1)+alpha*PN1(5);
            nCA2{p,q}(3,2)=nCA2{p,q}(3,2)+alpha*PN1(6);
            nCA2{p,q}(4,1)=nCA2{p,q}(4,1)+alpha*PN1(7);
        else
            nCA2{p,q}(1,3)=nCA2{p,q}(1,3)+alpha*PN2(1);
            nCA2{p,q}(1,4)=nCA2{p,q}(1,4)+alpha*PN2(2);
            nCA2{p,q}(2,2)=nCA2{p,q}(2,2)+alpha*PN2(3);
            nCA2{p,q}(2,3)=nCA2{p,q}(2,3)+alpha*PN2(4);
            nCA2{p,q}(3,1)=nCA2{p,q}(3,1)+alpha*PN2(5);
            nCA2{p,q}(3,2)=nCA2{p,q}(3,2)+alpha*PN2(6);
            nCA2{p,q}(4,1)=nCA2{p,q}(4,1)+alpha*PN2(7);
        end
    end
end

will you please explain why we taken (1,3)(1,4)....(4,1) and what is its significance

here pn1 and pn2 are pseudo noise sequences

[Walter Roberson]

will you please explain why we taken (1,3)(1,4)....(4,1) and what is its significance

I do not know.

Perhaps someone studied which locations in a block most reliably produced distinguishable changes that survive idwt -> dwt -> extract

[honey]

thank you

y1 = awgn(y1,10,'measured');

wii you explain in detail about awgn

[Walter Roberson]

The documentation page is https://www.mathworks.com/help/comm/ref/awgn.html . You could ask specific questions about that.

[honey]

Okk sir I would like to know about the wavread,auread and audioread commands and which one is the best among three

[Walter Roberson]

auread() used to exist, and it was a function to read NeXT/SUN (.au) sound files.

wavread() used to exist, and it was a function to read Microsoft WAVE (.wav) sound files.

Both auread() and wavread() have been removed from MATLAB, with audioread() now handling both kinds of files.

You should not use auread() or wavread() unless you need to use MATLAB R2012a or earlier; for any later release you should use audioread() or sometimes dsp.AudioFileReader https://www.mathworks.com/help/dsp/ref/dsp.audiofilereader-system-object.html

[honey]

thank you sir

did wavwrite be used in updated versions instead of audiowrite

and will you please explain the below each noise

ok_noise =0;        % select 1 for adding the particular disturbance to host audio
ok_filtering=0;
ok_cropping=0;
ok_resampling=0;
ok_requantization=0;
 
if ok_noise
    % Additional noise
    y1 = awgn(y1,10,'measured');
end
 
if ok_filtering
    % Filtering
    myfilter = ones(512,1);
    myfilter = myfilter/sum(myfilter);
    y1   = filter(myfilter,1,y1);
end
 
if ok_cropping
    % Cropping
    Lmin = round(L/11);
    Lmax = round(5*L/11);
    y1(1:1,1:Lmin)   = 0;
    y1(1:1,Lmax:end) = 0;
end
if ok_resampling
    % Resampling
    Fs_0 = Fs;
    Fs_1 = round(Fs/9);
    y1 = resample(y1,Fs_1,Fs_0);
    y1 = resample(y1,Fs_0,Fs_1);
    if length(y1)<L
        y1(end+L) = 0;
    end
    if length(y1)>L
        y1 = y1(1:L);
    end
end
 
if ok_requantization
    % Requantization
    bits_new = 8;
    wavwrite(y1,Fs,bits_new,'requantized_sound.wav');
    y1 = audioread('requantized_sound.wav');
end 

[Walter Roberson]

Yes wavwrite was replaced by audiowrite.

That code only has one noise, which is standard white Gaussian noise with signal to noise ratio 10

[honey]

i understand that only awgn is considering but i wanted to know about the remaining noises to and how they are working will you please explain those

[Walter Roberson]

There are no other noises in that code.

If you are asking what other possible kinds of noises there are in the world, then that would probably take a fair amount of research and time to answer.

[honey]

so those are just used for written purpose only is that correct

i didn't understand what is r---* in the below code

plot(qualityFactor,ssimValues,'r--*')

can you please explain it

[Walter Roberson]

I suspect that you might be confusing the code for filtering or cropping or requantization as being "noise".

r--* as a plot request says that the plot should be drawn in red ('r') and that it should use a dashed line ('--') and that each point should have a marker that is asterisk shaped ('*'). You can find more information in the documentation under the term "linespec"

[honey]

Is there any difference between reshaping and amplitude modulation while both are applied to a signal

[Walter Roberson]

Reshaping? Do you mean resampling?

[honey]

my question is reshaping and amplitude modulation can give same reults while applied to a signal

[Walter Roberson]

No. On MATLAB reshaping refers to taking the memory used to store an array and telling MATLAB to reinterpret the memory as a different set of dimensions. For example

reshape(1:12, 2, 3, 2)

uses the sequential memory used to store 1:12 and reinterprets it as 2 x 3 x 2. The values are not changed at all, and the total number of elements must be exactly the same afterwards, but now it is no longer 1 x 12 and is 2 x 3 x 2 instead.

Amplitude modulation is completely different, and refers to taking a carrier wave and imposing a signal on to it by modifying the amplitude of the carrier wave. The output is a higher frequency than the signal being carried, and the output is a vector.

[honey]

Thank you

[honey]

can any one of the dct and dwt technique is used for speech modulation for image watermarking

[Walter Roberson]

can any one of the dct and dwt technique is used for speech modulation for image watermarking

Yes. For the purpose of image watermarking, it does not matter how you create the stream of data that you use to watermark. If you want to convert a famous speech into Morse Code and use that, go ahead. If you want to use a picture of a baby goat, go ahead. If you want to use the historical record of high and low temperatures for your home town, go ahead. If you want to use dct or dwt to compress some audio and use that, go ahead.

[honey]

clc
clear all
close all
warning off
 
%% audio input
ok_classical=0;   % 1 selected 0 not-selected
ok_jazz=0;
ok_pop=1;
ok_looney=0;
 if ok_classical
[y,Fs] = audioread('classical.wav');
 end
 if ok_jazz
[y,Fs] = audioread('jazz.wav');
 end
 if ok_pop
[y,Fs] = audioread('pop.wav');
 end
 if ok_looney
[y,Fs] = audioread('loopyMusic.wav');
 end
x1=y(1:262144);
x1=reshape(x1,1,262144);
x=reshape(x1,512,512);
Key=7;
[L1, L2]=size(x1);
%% image input 
 
img  = imread('CW32.jpg'); %Get the input image
I = rgb2gray(img);
original_img=im2bw(I);
w1=im2bw(I);%Convert to grayscale image
 
%% 3 level of DWT decomposition with HAAR wavelet
 
[L,H,V,D]=dwt2(x,'haar','sym');
[CL,CH,CV,CD]=dwt2(H,'haar','sym');
[CA1,CA2,CA3,CA4]=dwt2(CH,'haar','sym');
for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
%% DCT of sub-band block CH->CA2
 
newCA2=mat2cell(CA2,a,b);
d=zeros(64,64);
nCA2=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nCA2{i,j}=dct2(newCA2{i,j});
    end
end
%% DCT of sub-band block CV->CV2
 
[CV1,CV2,CV3,CV4]=dwt2(CV,'haar','sym');
 
newCV2=mat2cell(CV2,a,b);
d=zeros(64,64);
nCV2=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nCV2{i,j}=dct2(newCV2{i,j});
    end
end
%% DCT of sub-band block VH->VH2
 
[VL,VH,VV,VD]=dwt2(V,'haar','sym');
[VH1,VH2,VH3,VH4]=dwt2(VH,'haar','sym');
newVH2=mat2cell(VH2,a,b);
d=zeros(64,64);
nVH2=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nVH2{i,j}=dct2(newVH2{i,j});
    end
end
%% dct OF SUB-band VV->VV2
 
[VV1,VV2,VV3,VV4]=dwt2(VV,'haar','sym');
newVV2=mat2cell(VV2,a,b);
d=zeros(64,64);
nVV2=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nVV2{i,j}=dct2(newVV2{i,j});
    end
end
%% UNCORRELATED PSEUDO RANDOM SEQUENCE GENERATOR
 
G=Key;
%Generation of first m-sequence using generator polynomial [45]
sd1 =[0 0 0 0 1];             % Initial state of Shift register
PN1=[];                       % First m-sequence
for j=1:G
    PN1=[PN1 sd1(5)];
    if sd1(1)==sd1(4)
        temp1=0;
    else temp1=1;
    end
    sd1(1)=sd1(2);
    sd1(2)=sd1(3);
    sd1(3)=sd1(4);
    sd1(4)=sd1(5);
    sd1(5)=temp1;
end
sd2 =[0 1 0 0 1];             % Initial state of Shift register
PN2=[];                       % Second m-sequence
for j=1:G
    PN2=[PN2 sd2(5)];
    if sd2(1)==sd2(2)
        temp1=0;
    else temp1=1;
    end
    if sd2(4)==temp1
        temp2=0;
    else temp2=1;
    end
    if sd2(5)==temp2
        temp3=0;
    else temp3=1;
    end
    sd2(1)=sd2(2);
    sd2(2)=sd2(3);
    sd2(3)=sd2(4);
    sd2(4)=sd2(5);
    sd2(5)=temp3;
end
%% WATERMARKING COEFFICIENT OR WEIGHT ALPHA
 
alpha=0.1;
%% WATERMARK BIT ENCODING ALGORITHM IN SELECTED/MID-BAND COEFFICIENTS OF DCT TRANSFORMED BLOCKS ONLY
%% FOR BLOCK CA2
 
for p=1:16
    for q=1:16
        if w1(p,q)==0
            nCA2{p,q}(1,3)=nCA2{p,q}(1,3)+alpha*PN1(1);
            nCA2{p,q}(1,4)=nCA2{p,q}(1,4)+alpha*PN1(2);
            nCA2{p,q}(2,2)=nCA2{p,q}(2,2)+alpha*PN1(3);
            nCA2{p,q}(2,3)=nCA2{p,q}(2,3)+alpha*PN1(4);
            nCA2{p,q}(3,1)=nCA2{p,q}(3,1)+alpha*PN1(5);
            nCA2{p,q}(3,2)=nCA2{p,q}(3,2)+alpha*PN1(6);
            nCA2{p,q}(4,1)=nCA2{p,q}(4,1)+alpha*PN1(7);
        else
            nCA2{p,q}(1,3)=nCA2{p,q}(1,3)+alpha*PN2(1);
            nCA2{p,q}(1,4)=nCA2{p,q}(1,4)+alpha*PN2(2);
            nCA2{p,q}(2,2)=nCA2{p,q}(2,2)+alpha*PN2(3);
            nCA2{p,q}(2,3)=nCA2{p,q}(2,3)+alpha*PN2(4);
            nCA2{p,q}(3,1)=nCA2{p,q}(3,1)+alpha*PN2(5);
            nCA2{p,q}(3,2)=nCA2{p,q}(3,2)+alpha*PN2(6);
            nCA2{p,q}(4,1)=nCA2{p,q}(4,1)+alpha*PN2(7);
        end
    end
end
 
for i=1:16
    for j=1:16
        newCA2{i,j}=idct2(nCA2{i,j});
    end
end
CA2=cell2mat(newCA2);
%% FOR BLOCK CV2
 
for p=1:16
    for q=1:16
        if w1(p,16+q)==0
            nCV2{p,q}(1,3)=nCV2{p,q}(1,3)+alpha*PN1(1);
            nCV2{p,q}(1,4)=nCV2{p,q}(1,4)+alpha*PN1(2);
            nCV2{p,q}(2,2)=nCV2{p,q}(2,2)+alpha*PN1(3);
            nCV2{p,q}(2,3)=nCV2{p,q}(2,3)+alpha*PN1(4);
            nCV2{p,q}(3,1)=nCV2{p,q}(3,1)+alpha*PN1(5);
            nCV2{p,q}(3,2)=nCV2{p,q}(3,2)+alpha*PN1(6);
            nCV2{p,q}(4,1)=nCV2{p,q}(4,1)+alpha*PN1(7);
        else
            nCV2{p,q}(1,3)=nCV2{p,q}(1,3)+alpha*PN2(1);
            nCV2{p,q}(1,4)=nCV2{p,q}(1,4)+alpha*PN2(2);
            nCV2{p,q}(2,2)=nCV2{p,q}(2,2)+alpha*PN2(3);
            nCV2{p,q}(2,3)=nCV2{p,q}(2,3)+alpha*PN2(4);
            nCV2{p,q}(3,1)=nCV2{p,q}(3,1)+alpha*PN2(5);
            nCV2{p,q}(3,2)=nCV2{p,q}(3,2)+alpha*PN2(6);
            nCV2{p,q}(4,1)=nCV2{p,q}(4,1)+alpha*PN2(7);
        end
    end
end
 
for i=1:16
    for j=1:16
        newCV2{i,j}=idct2(nCV2{i,j});
    end
end
CV2=cell2mat(newCV2);
%% FOR BLOCK VH2
 
for p=1:16
    for q=1:16
        if w1(p+16,q)==0
            nVH2{p,q}(1,3)=nVH2{p,q}(1,3)+alpha*PN1(1);
            nVH2{p,q}(1,4)=nVH2{p,q}(1,4)+alpha*PN1(2);
            nVH2{p,q}(2,2)=nVH2{p,q}(2,2)+alpha*PN1(3);
            nVH2{p,q}(2,3)=nVH2{p,q}(2,3)+alpha*PN1(4);
            nVH2{p,q}(3,1)=nVH2{p,q}(3,1)+alpha*PN1(5);
            nVH2{p,q}(3,2)=nVH2{p,q}(3,2)+alpha*PN1(6);
            nVH2{p,q}(4,1)=nVH2{p,q}(4,1)+alpha*PN1(7);
        else
            nVH2{p,q}(1,3)=nVH2{p,q}(1,3)+alpha*PN2(1);
            nVH2{p,q}(1,4)=nVH2{p,q}(1,4)+alpha*PN2(2);
            nVH2{p,q}(2,2)=nVH2{p,q}(2,2)+alpha*PN2(3);
            nVH2{p,q}(2,3)=nVH2{p,q}(2,3)+alpha*PN2(4);
            nVH2{p,q}(3,1)=nVH2{p,q}(3,1)+alpha*PN2(5);
            nVH2{p,q}(3,2)=nVH2{p,q}(3,2)+alpha*PN2(6);
            nVH2{p,q}(4,1)=nVH2{p,q}(4,1)+alpha*PN2(7);
        end
    end
end
 
for i=1:16
    for j=1:16
        newVH2{i,j}=idct2(nVH2{i,j});
    end
end
VH2=cell2mat(newVH2);
%% FOR BLOCK VV2
 
for p=1:16
    for q=1:16
        if w1(p+16,q+16)==0
            nVV2{p,q}(1,3)=nVV2{p,q}(1,3)+alpha*PN1(1);
            nVV2{p,q}(1,4)=nVV2{p,q}(1,4)+alpha*PN1(2);
            nVV2{p,q}(2,2)=nVV2{p,q}(2,2)+alpha*PN1(3);
            nVV2{p,q}(2,3)=nVV2{p,q}(2,3)+alpha*PN1(4);
            nVV2{p,q}(3,1)=nVV2{p,q}(3,1)+alpha*PN1(5);
            nVV2{p,q}(3,2)=nVV2{p,q}(3,2)+alpha*PN1(6);
            nVV2{p,q}(4,1)=nVV2{p,q}(4,1)+alpha*PN1(7);
        else
            nVH2{p,q}(1,3)=nVV2{p,q}(1,3)+alpha*PN2(1);
            nVV2{p,q}(1,4)=nVV2{p,q}(1,4)+alpha*PN2(2);
            nVV2{p,q}(2,2)=nVV2{p,q}(2,2)+alpha*PN2(3);
            nVV2{p,q}(2,3)=nVV2{p,q}(2,3)+alpha*PN2(4);
            nVV2{p,q}(3,1)=nVV2{p,q}(3,1)+alpha*PN2(5);
            nVV2{p,q}(3,2)=nVV2{p,q}(3,2)+alpha*PN2(6);
            nVV2{p,q}(4,1)=nVV2{p,q}(4,1)+alpha*PN2(7);
        end
    end
end
 
for i=1:16
    for j=1:16
        newVV2{i,j}=idct2(nVV2{i,j});
    end
end
VV2=cell2mat(newVV2);
%% IDWT AND IDCT TO GET MODIFIED COEFFICIENTS AND FORM THE WATERMARKED AUDIO 
 
CH=idwt2(CA1,CA2,CA3,CA4,'haar','sym');
CV=idwt2(CV1,CV2,CV3,CV4,'haar','sym');
H=idwt2(CL,CH,CV,CD,'haar','sym');
VH=idwt2(VH1,VH2,VH3,VH4,'haar','sym');
VV=idwt2(VV1,VV2,VV3,VV4,'haar','sym');
V=idwt2(VL,VH,VV,VD,'haar','sym');
newx=idwt2(L,H,V,D,'haar','sym');
y1=reshape(newx,1,512*512);
SNR=snr(y1,x1)
y2=reshape(newx,1,512*512);
L=length(y1);
%% DISTURBANCE ADDED TO THE WATERMARKED AUDIO
 
ok_noise =0;        % select 1 for adding the particular disturbance to host audio

 
if ok_noise
    % Additional noise
    y1 = awgn(y1,10,'measured');
end
 

%% WATERMARK EXTRACTION ALGORITHM
 
newx=reshape(y1,512,512);
[A,B,C,D]=dwt2(newx,'haar','sym');
[B1,B2,B3,B4]=dwt2(B,'haar','sym');
[B21,B22,B23,B24]=dwt2(B2,'haar','sym');
[B31,B32,B33,B34]=dwt2(B3,'haar','sym');
[C1,C2,C3,C4]=dwt2(C,'haar','sym');
[C21,C22,C23,C24]=dwt2(C2,'haar','sym');
[C31,C32,C33,C34]=dwt2(C3,'haar','sym');
%% DCT OF SUB-BANDS
 
for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
newB22=mat2cell(B22,a,b);
d=zeros(64,64);
nB22=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nB22{i,j}=dct2(newB22{i,j});
    end
end
%% EXTRACTION OF WATERMARK BIT BY COMPARISON OF CORRELATION
 
for p=1:16
    for q=1:16
        if corr([nB22{p,q}(1,3) nB22{p,q}(1,4) nB22{p,q}(2,2) nB22{p,q}(2,3) nB22{p,q}(3,1) nB22{p,q}(3,2) nB22{p,q}(4,1)]',PN1(1:7)')>=corr([nB22{p,q}(1,3) nB22{p,q}(1,4) nB22{p,q}(2,2) nB22{p,q}(2,3) nB22{p,q}(3,1) nB22{p,q}(3,2) nB22{p,q}(4,1)]',PN2(1:7)')
            w2(p,q)=0;
        else
            w2(p,q)=1;
        end
    end
end
%% DCT OF SUB-BANDS
 
for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
newB32=mat2cell(B32,a,b);
d=zeros(64,64);
nB32=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nB32{i,j}=dct2(newB32{i,j});
    end
end
%% EXTRACTION OF WATERMARK BIT BY COMPARISON OF CORRELATION
 
for p=1:16
    for q=1:16
        if corr([nB32{p,q}(1,3) nB32{p,q}(1,4) nB32{p,q}(2,2) nB32{p,q}(2,3) nB32{p,q}(3,1) nB32{p,q}(3,2) nB32{p,q}(4,1)]',PN1(1:7)')>=corr([nB32{p,q}(1,3) nB32{p,q}(1,4) nB32{p,q}(2,2) nB32{p,q}(2,3) nB32{p,q}(3,1) nB32{p,q}(3,2) nB32{p,q}(4,1)]',PN2(1:7)')
            w2(p,q+16)=0;
        else
            w2(p,q+16)=1;
        end
    end
end
 
%% DCT OF SUB-BAND
 
for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
newC22=mat2cell(C22,a,b);
d=zeros(64,64);
nC22=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nC22{i,j}=dct2(newC22{i,j});
    end
end
%% EXTRACTION OF WATERMARK BIT BY COMPARISON OF CORRELATION
 
for p=1:16
    for q=1:16
        if corr([nC22{p,q}(1,3) nC22{p,q}(1,4) nC22{p,q}(2,2) nC22{p,q}(2,3) nC22{p,q}(3,1) nC22{p,q}(3,2) nC22{p,q}(4,1)]',PN1(1:7)')>=corr([nC22{p,q}(1,3) nC22{p,q}(1,4) nC22{p,q}(2,2) nC22{p,q}(2,3) nC22{p,q}(3,1) nC22{p,q}(3,2) nC22{p,q}(4,1)]',PN2(1:7)')
            w2(p+16,q)=0;
        else
            w2(p+16,q)=1;
        end
    end
end
%% DCT OF SUB-BAND
 
for i=1:16
    a(i)=4;
end
for i=1:16
    b(i)=4;
end
newC32=mat2cell(C32,a,b);
d=zeros(64,64);
nC32=mat2cell(d,a,b);
for i=1:16
    for j=1:16
        nC32{i,j}=dct2(newC32{i,j});
    end
end
%% EXTRACTION OF WATERMARK BIT BY COMPARISON OF CORRELATION
 
for p=1:16
    for q=1:16
        if corr([nC32{p,q}(1,3) nC32{p,q}(1,4) nC32{p,q}(2,2) nC32{p,q}(2,3) nC32{p,q}(3,1) nC32{p,q}(3,2) nC32{p,q}(4,1)]',PN1(1:7)')>=corr([nC32{p,q}(1,3) nC32{p,q}(1,4) nC32{p,q}(2,2) nC32{p,q}(2,3) nC32{p,q}(3,1) nC32{p,q}(3,2) nC32{p,q}(4,1)]',PN2(1:7)')
            w2(p+16,q+16)=0;
        else
            w2(p+16,q+16)=1;
        end
    end
end
sbr=w2;
figure
imshow(logical(w2))
%% Quality check
q=max(imabsdiff(y2,reshape(x,1,512*512)))
BER=biterr(original_img,sbr)
ssimval = ssim(uint8(sbr),uint8(original_img))
ps=psnr(uint8(original_img),uint8(sbr)) % Must be greater than 35 dB
ssimValues = zeros(1,10);
qualityFactor = 10:10:100;
for i = 1:10
    imwrite(I,'sbr.jpg','jpg','quality',qualityFactor(i));
    ssimValues(i) = ssim(imread('sbr.jpg'),I);
end
figure
plot(qualityFactor,ssimValues,'r--*')
xlabel(' Copression Quality Factor');
ylabel(' SSIM value');
%% END %%

can u please explain in this code why is it necessary to use both DCT and DWT and also why only HAAR in DWT

[Walter Roberson]

can u please explain in this code why is it necessary to use both DCT and DWT

You should be asking the author of the code questions like that.

why only HAAR in DWT

'haar' has good time localization, able to deal with sudden changes in value, which makes it more suitable for use in audio.

[honey]

thank you

Answer 2

[the cyclist]

Use the reshape command.