LOCALIZATION OF SOUND/TONE (3D SOUND)

i am currently working on 3D sound. i have employed the alogorithm and formula from c.phillps brown and richard Duda research paper. i am unable to make prominant difference between front and back discrimination of sound. also tell me some method to externalize the sound. please help me in this

[Image Analyst]

This forum is for help with MATLAB code, not for explaining papers or how published algorithms work. Have you written the MATLAB code yet?

[Walter Roberson]

I believe the paper is doi 10.1.1.21.8822

[muhammad tariq]

yes the same paper....i have posted the code on this thread as well.....can u suggest some correction/addition

thanks

Answer 1

[Daniel Shub]

Note to be rude, but rather pragmatic, you can't. In the absence of head movement and reverberation front-back confusions are frequent, and distance perception, somewhat related to externalization, is poor.

[muhammad tariq]

thanks fr reply,

but in a audio file like virtual barbershop and other virtual sounds...you can feel the sound coming from front or back

[Daniel Shub]

Yes, but virtual barbershop (http://www.youtube.com/watch?v=IUDTlvagjJA) relies on a situation in which you have very strong priors. It doesn't work in general.

[muhammad tariq]

can you kindly tel me what do you mean by stong priors

[Daniel Shub]

I mean priors as in Bayesian priors. You expect the clippers to be at the back of your head, so you hear them there. If you were to reprocess the sounds in virtual barbershop flipping the front to the back and vice versa, I bet the percept would not be as strong. The same can be said of the surround sound in movie theaters (and homes), rarely are we in the optimal position for listening, so the cues are wrong, but we hear the sounds in the correct place.

Answer 2

[muhammad tariq]

the code is here.in this code ITD,ILD and PINNA MODEL (fr filtering) have been implemented in order to discriminate front and back side sound/tone...but failed to achieve the required results.furthermore m unable to achieve externalization(out of head) of sound. i have made a GUI and have taken recorded audio files.i m changing the location(changing azimuth angle of arrival of sound) on GUI and observing the sound via headphones. what i want to do is when i take the sound to the back of head i should feel the sound in the headphones as coming from back side of head.i have done the work in which the direction of sound on front half side of head can be observed....but from the backside the feeling is quite weak.any help in this regard will be highly appreciated.

function [y] = hrtfsphere(a,Dk,az,elev,snd,fs)
    c = 334; %Speed of sound in m/s
    theta_left = -90*pi/180; 
    theta_right = 90*pi/180;

      wo = c/a;

      az = az*pi/180;

      if (elev > 90)
          elev = 90;
      elseif (elev < -90)
          elev  =-90;
      end

      elev = elev*pi/180;    

      Ak = [1 5 5 5 5];        %coefficeint values for pinna model
      Bk = [2 4 7 11 13];
      g = [0.5 -1 .5 -0.25 0.25];

      snd1 = snd(:,1);
      if (size(snd,2) == 2)
        snd2 = snd(:,2);
      else 
        snd2 = snd1;
      end

%implementing ear pinna filtering as FIR filter

%pinna model          

Tk = cos(az/2) * Ak .* sin(Dk*(pi/2-elev)) + Bk;

      coeffs = zeros(1,max(ceil(Tk))+2);
      for p = 1:5
        c = floor(Tk(p));
        r = mod(Tk(p),c);
  %finding coefficient(weights) for FIR filtr   
        coeffs(c) = coeffs(c) + g(p)*(1-r);    
        coeffs(c+1) = coeffs(c+1) + g(p)*r;    
      end
      Tk = [1 coeffs];
      snd1 = filter(Tk,1,snd1); %implementin FIR filter
      snd2 = filter(Tk,1,snd2);

      al = (1.05 + 0.95*cos((az-theta_left)*18/15));
      ar = (1.05 + 0.95*cos((az-theta_right)*18/15));

      %calculating time delay b/w the two ears
      dtl = 0;  
      if (0 <= abs(az-theta_left) < pi/2)
          dtl = -a/c*cos(az-theta_left);
      elseif (pi/2 <= abs(az-theta_left)< pi)
          dtl = a/c*(abs(az-theta_left)-pi/2);
      end
      dtl = dtl + a/c;

      dtr = 0;
      if (0 <= abs(az-theta_right) < pi/2)
          dtr = -a/c*cos(az-theta_right);
      elseif (pi/2 <= abs(az-theta_right)< pi)
          dtr = a/c*(abs(az-theta_right)-pi/2);
      end
      dtr = dtr + a/c;

      x1 = fft(snd1);
      x1 = x1(1:floor(length(x1)/2))';
      x2 = fft(snd2);
      x2 = x2(1:floor(length(x2)/2))';

      f = fs*(0:length(x1)-1)/length(x1);
      w = 2*pi*f; 

      Hl = ( 1 + ( 1i*al*w / (2*wo))) ./ (1+1i*w/(2*wo)) .* exp(-1i*w*dtr);
      Hr = ( 1 + ( 1i*ar*w / (2*wo))) ./ (1+1i*w/(2*wo)) .* exp(-1i*w*dtl);

      yl = x1.*Hl;
      yr = x2.*Hr;

      yl(length(yl) + 1) = 0;
      yl = [yl conj(fliplr(yl(2:length(yl)-1)))];   

      yr(length(yr) + 1) = 0;
      yr = [yr conj(fliplr(yr(2:length(yr)-1)))];   

      y(:,1) = fliplr(ifft(yl(1:length(yl))));
      y(:,2) = fliplr(ifft(yr(1:length(yr))));
      y = y./(1.5*max(max(y)));

subplot(3,3,[3 6]), plot(y);

axis([0 20 -1 1]);

legend ('Left Channel / Ear' , 'Right Channel / Ear ')

title('Output Result on Channels');
xlabel('Samples');
ylabel('Amplitude');
grid on;

end

Answer 3

[Image Analyst]

If you're interested in visualizing sound and determining where it came from and its intensity at each location you can use an acoustic camera, such as this one: http://www.acoustic-camera.com/en/acoustic-camera-en. It produces an image of sound intensity - see the web site for examples. If you had several of those you could perhaps get a 3D volumetric image of where sound originates.

[muhammad tariq]

thanks for reply...

actually i have made a GUI and have taken recorded audio files.i m changing the location(changing azimuth angle of arrival of sound) on GUI and observing the sound via headphones.

what i want to do is when i take the sound to the back of head i should feel the sound in the headphones as coming from back side of head.i have done the work in which the direction of sound on front half side of head can be observed....but from the backside the feeling is quite weak...please help out in ths regard

[Image Analyst]

I can't help you. I know nothing about audio processing.

[muhammad tariq]

can u who tel me anybdy who knws audio processing

Answer 4

[jhosmar solarte]

Buena noche amigo.

Será que alguien me podría regalar el código completo por favor.

[Walter Roberson]

Not likely . We are not a code writing service . If you encounter errors in your own implementation then we will help you find the problem .