How can I divide an audio signal into overlap frames and take FFT to the signal?

Hi,

I need to divide an audio signal into overlap frames of duration = 0.025 sec each, with frame step of 0.01 sec. and multiply each frame by the Hamming window and then take the FFT to the windowed signal......can some one help me please?

[s,Fs] = audioread('my file.wav');

%%%fs = 44100 hz
n=512;              %Number of FFT points

Tf=0.025;           %Frame duration in seconds

N=Fs*Tf;            %Number of samples per frame

Ts=0.01;            %Frame step in seconds

FrameStep=Fs*Ts;    %Frame step in samples

[Zhi Zhang]

excuse me, sir, have you already solved that?

Answer 1

[Walter Roberson]

Use buffer()

[nada fady]

Thank you for replay sir,

I tried this

%Read the data to the MATLAB using audioread.

[s,fs] = audioread('C:\Users\Fady\Desktop\NADA MSc\h.wav');

N=length(s)                     %total num of samples

ts=0.01                         %Frame step in seconds

frame_step=floor(ts*fs)         %Frame step in samples

frame_duration=0.025        %Frame duration in seconds

frame_length=ceil(frame_duration*fs)  %Number of samples per frame

frame_num=ceil(N/frame_step)

y= buffer (N,frame_length,frame_step)

but I get

y=0
0
0
0
0
.
.
.
.
 155614
0
0
0
0
.
.
.

I don't know what is the wrong

%note: 155614 is my data length.

[Walter Roberson]

for K = 1 : size(s,2)
  y{K} = buffer(s(:,K),frame_length,frame_step);
end

This code takes into account the possibility that your audio file has multiple channels.

[Gabriel Hurtado]

When yall say frame step, is that referring to the overlap of the frame? i.e. frame duration is the length of a segment, while frame step would refer to where the next segment begins?

If frame duration is 25 ms, and frame step is 10 ms, then the first segment has samples between 0 - 25 ms and the next has samples between 10 - 35 ms and the following has samples between 20 - 45 ms and so on? That's the idea of this code right? Just want to make sure.

Answer 2

[Reza Safdari]

You can use enframe.m function from voicebox, a speech processing toolbox for MATLAB. For more detailed information refer to the following link: http://www.ee.ic.ac.uk/hp/staff/dmb/voicebox/doc/voicebox/enframe.html#_top.

Answer 3

[Manolis Michailidis]

There is a function called spectrogram which does what you want , see the related documentation http://www.mathworks.com/help/signal/ref/spectrogram.html?searchHighlight=spectrogram

[Parul Sahi]

Sir

how does the spectrogram function calculate the number of frames into which the signal will be divided given all the arguments like window size, hop length, NFFT points, window type?

Answer 4

[Pooja Patel]

you can split by number of samples or time duration or frame