How can I divide an audio signal into frames and write the frames into a matrix?

Hi all!

I need to divide an audio signal into frames of length = 128 samples each, then multiply each frame by the Hamming window and write the windowed frames into a matrix where each row of this matrix contains a single frame.

This is what I did so far:(I'm stuck at the for loop!)

[input, fs] = wavread('.....'); % input= <15360x1 double> , fs= 22050
N=15350;
L = length(input)./fs;          % time in seconds -> L=0.96
w_length = 128;                 % window length
w = hamming(w_length);          % window type - hamming
% windowing and matrix creation
for 1:w_length:N                
    .
    .
end

Thank you in advance!

[Shashank R]

@Geoff Hayes, I tried the same thing and it kept showing me the error when I tried to apply the Hamming Window. Could you please help me out. I'm new to Matlab and also as David asked about recovering the signal through concatenation.

Thanks in Advance

[Geoff Hayes]

Shashank - which error? Please copy and paste the full error message and describe what you are doing. You may want to post this as a new question if it is significantly different from this one.

Answer 1

[Geoff Hayes]

David - what is N? Is that the assumed length of the input?

Probably one of the first things that you want to do is to allocate memory to the matrix to store each frame/block of data. You know that the Hamming window size is 128 (which will be the number of columns in the matrix), and you need to determine the number of rows

 % determine the number of rows/frames
 m         = length(input);
 numFrames = floor((m-1)/w_length)+1;

 % allocate memory to the frame matrix
 frameData = zeros(numFrames,w_length);

Now, extract the frames from the input data

 for k=1:numFrames
     startAtIdx = (k-1)*w_length+1;
     if k~=numFrames
         frameData(k,:) = input(startAtIdx:startAtIdx+w_length-1);
     else
         % handle this case separately in case the number of input samples
         % does not divide evenly by the window size
         frameData(k,1:m-startAtIdx+1) = input(startAtIdx:end);
     end
  end

All that is missing is to apply the Hamming window...which I will leave for you. Try the above and see what happens!

[David]

Hi Geoff Hayes and thanks for the prompt reply!

Sorry that the question wasn't clear enough. Your code answered my question;

Yes N= input length. I needed to divide the N= 15350 by frames of 128 in length. so 15350/128=120 thus [128 col x 120 row]. I tested your code and works perfectly!

If it's not much to ask.. how can I reverse the process? by that I mean: concatenating the 120 rows into a single output signal vector, so I can playback the audio file.

Thanks again Geoff Hayes!

[Geoff Hayes]

Hi David - just use the reshape function to convert the matrix into a vector e.g.

 A = [1 2 3 4; 5 6 7 8; 9 10 11 12]

 reshape(A',1,numel(A))

Note the transpose of A as the first input to the above function - without it, you won't get the correct solution!

[Shashank R]

@David, Could you please help out with the procedure of application of the Hamming Window and concatenation?

Thank You

[Sumayyah Hai]

@Geoff Hi! I was wondering how you would implement the hamming window on the framed matrix. What will you use as input argument for the: w=hamming(L) command? The number of samples in one frame, 128, or the number of frames (rows)? And after getting the w vector how would you implement it on the frameData matrix?

[mugambi ndwiga]

@Sumayyah Just multiply w with the frame . i.e.

framedata .* w