Forums

binaural 3d sound localization in reverberant room

Started by riz May 19, 2007
Which algorithm is most effective in binaural 3d sound localization in a
reverberant environment(using only two microphones)?.i have tried to
localize sound in 3d (reverberant environment) using hrtfs i.e convolving
the left and right microphones inputs (each of 512 samples) with 710
available inverse hrtfs and then correlating them.it doesnot give good
results especially the elevation estimation has high deviation.any clues
?


_____________________________________
Do you know a company who employs DSP engineers?  
Is it already listed at http://dsprelated.com/employers.php ?
Why the heck do you need hrtf ?

Are you trying to fit your mics on a dummy head ?
Just give me a break...


On May 19, 12:30 pm, "riz" <rizwan....@gmail.com> wrote:
> Which algorithm is most effective in binaural 3d sound localization in a > reverberant environment(using only two microphones)?.i have tried to > localize sound in 3d (reverberant environment) using hrtfs i.e convolving > the left and right microphones inputs (each of 512 samples) with 710 > available inverse hrtfs and then correlating them.it doesnot give good > results especially the elevation estimation has high deviation.any clues > ? > > _____________________________________ > Do you know a company who employs DSP engineers? > Is it already listed athttp://dsprelated.com/employers.php?
yes,the mics are on a dummy head

>Why the heck do you need hrtf ? > >Are you trying to fit your mics on a dummy head ? >Just give me a break... > > >On May 19, 12:30 pm, "riz" <rizwan....@gmail.com> wrote: >> Which algorithm is most effective in binaural 3d sound localization in
a
>> reverberant environment(using only two microphones)?.i have tried to >> localize sound in 3d (reverberant environment) using hrtfs i.e
convolving
>> the left and right microphones inputs (each of 512 samples) with 710 >> available inverse hrtfs and then correlating them.it doesnot give good >> results especially the elevation estimation has high deviation.any
clues
>> ? >> >> _____________________________________ >> Do you know a company who employs DSP engineers? >> Is it already listed athttp://dsprelated.com/employers.php? > > >
_____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
On May 20, 5:30 am, "riz" <rizwan....@gmail.com> wrote:
> Which algorithm is most effective in binaural 3d sound localization in a > reverberant environment(using only two microphones)?.i have tried to > localize sound in 3d (reverberant environment) using hrtfs i.e convolving > the left and right microphones inputs (each of 512 samples) with 710 > available inverse hrtfs and then correlating them.it doesnot give good > results especially the elevation estimation has high deviation.any clues > ? > > _____________________________________ > Do you know a company who employs DSP engineers? > Is it already listed athttp://dsprelated.com/employers.php?
I don't see how you can with 2 mics. You would need to re-orientate them otherwise how do you get the z component? Besides that, you have front-back ambiguity at the line that connects the two mics. Suggest you use 4 mics and use the Spherical Intersection method. The difficult bit is estimating delay in a reverberant room. Try the Hanan- Thomson method of Generalised cross correlation. Wang King
It is assumed that we are getting left and right signals at the two
microphones as original signal that is passed through a specific HRTF
filter(both left and right)and received at two microphones.KEMAR HRTF
measurements are 710 measurements of HRTFs at different angle of azimuth
and elevation.So if we perform the convolution of left and right
microphone signals with inverse of original HRTFs(left and right),we must
get the two signals that are highly correlated.This particular HRTF which
results in highest correlation will correspond to a specific azimuth and
elevation in 3D.

>On May 20, 5:30 am, "riz" <rizwan....@gmail.com> wrote: >> Which algorithm is most effective in binaural 3d sound localization in
a
>> reverberant environment(using only two microphones)?.i have tried to >> localize sound in 3d (reverberant environment) using hrtfs i.e
convolving
>> the left and right microphones inputs (each of 512 samples) with 710 >> available inverse hrtfs and then correlating them.it doesnot give good >> results especially the elevation estimation has high deviation.any
clues
>> ? >> >> _____________________________________ >> Do you know a company who employs DSP engineers? >> Is it already listed athttp://dsprelated.com/employers.php? > >I don't see how you can with 2 mics. You would need to re-orientate >them otherwise how do you get the z component? Besides that, you have >front-back ambiguity at the line that connects the two mics. Suggest >you use 4 mics and use the Spherical Intersection method. The >difficult bit is estimating delay in a reverberant room. Try the Hanan- >Thomson method of Generalised cross correlation. > >Wang King > >
_____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
On 19 mai, 19:30, "riz" <rizwan....@gmail.com> wrote:
> Which algorithm is most effective in binaural 3d sound localization in a > reverberant environment(using only two microphones)?.i have tried to > localize sound in 3d (reverberant environment) using hrtfs i.e convolving > the left and right microphones inputs (each of 512 samples) with 710 > available inverse hrtfs and then correlating them.it doesnot give good > results especially the elevation estimation has high deviation.any clues > ? > > _____________________________________ > Do you know a company who employs DSP engineers? > Is it already listed athttp://dsprelated.com/employers.php?
Hi, have a look to Blind Source Separation (BSS) methods, in particular in sparsity-based methods. You can easily use them as Source Localization techniques, as shown in [1]. Concerning the sparse BSS methods, I recommand you to read [2-7] (the list is not exhaustive). Have also a look to the methods using CASA (for example, the papers published here: http://www.icassp2006.org/Papers/PublicSessionIndex3.asp?SessionID=1155) Matthieu PS: my papers [4-5] are accessible form my webpage: www.ast.obs-mip.fr/puigt. [3] is accessible from www.ast.obs-mip.fr/deville [1] J.Mouba, S. Marchand, A source localization/separation/ respatialization system based on unsupervised classification of interaural cues, Proc. of DAFx 2006, Montreal, Canada, September 18-20, 2006. [2] O. Yilmaz, S. Rickard, Blind Separation of Speech Mixtures via Time-Frequency Masking, IEEE Transactions on Signal Processing, Vol. 52, No. 7, pages 1830-1847, July 2004. [3] F. Abrard, Y. Deville, A time-frequency blind signal separation method applicable to underdetermined mixtures of dependent sources, Signal Processing, vol. 85, issue 7, pp. 1389-1403, July 2005. [4] Y. Deville, M. Puigt, Temporal and time-frequency correlation- based blind source separation methods. Part I : determined and underdetermined linear instantaneous mixtures, Signal Processing, vol. 87, no. 3, pp. 374-407, March 2007. [5] M. Puigt, Y. Deville, Time-frequency ratio-based blind separation methods for attenuated and time-delayed sources, Mechanical Systems and Signal Processing, vol. 19, pp. 1348-1379, 2005. [6] S. Arberet, R. Gribonval, F. Bimbot, A Robust Method to Count and Locate Audio Sources in a Stereophonic Linear Anechoic Mixture, to appear in Proc. IEEE Intl. Conf. Acoust. Speech Signal Process (ICASSP'07)