September Meeting Notice

September 2004 Meeting Notice

The two standard methods for capturing live spatial sound with microphone arrays attempt either to reconstruct the incoming wave field or to reproduce the sound pressure at the eardrums. The various surround-sound systems usually use loudspeakers to provide approximate wave field synthesis, and  the effect of the listener's head is automatically accounted for by immersing the listener in the sound field.  Unfortunately, exact wave field reconstruction requires a very large number of channels.  Binaural or dummy-head recordings, which need only two channels, usually use headphones to provide separate signals directly to the ears.  Here the effect of the listener's head is approximately accounted for by the corresponding dummy head.  Unfortunately, binaural recordings do not account for the dynamic effects that occur when listeners turn their heads, and the sonic world turns whenever the listener turns.

 We describe a new approach that provides a different tradeoff between bandwidth requirements and spatial fidelity.  Our new method -- called MTB for Motion-Tracked Binaural --  is a generalization of binaural recordings that uses more than two channels so that it can capture the dynamic effects generated by the listener's motion and stabilize the acoustic field.  The result, which is best heard over headphones, is highly realistic reproduction of spatial sound with modest bandwidth requirements.  The method requires interpolation between microphone signals, and benefits from customization for individual listeners.  Interpolation and customization methods will be described, and a demonstration of the system will be given.

Richard O. Duda received his B.S. and M.S. degrees in Engineering from UCLA in 1958 and 1959, respectively, and his Ph.D. in Electrical
Engineering from MIT in 1962. He was in the Artificial Intelligence Center at SRI International from 1962 to 1980, serving as a visiting professor at the University of Texas at Austin in the 1973/74 academic year. He was in the Laboratory for AI Research at Fairchild Semiconductor from 1980 to 1983,
was with Syntelligence from 1983 to 1988, and was a Professor of Electrical Engineering at San Jose State University from 1988 to 2001. Since then he has been performing research on spatial sound in the CIPIC Interface Laboratory at UC Davis.

 Dr. Duda is the co-author with Peter E. Hart and David G. Stork of Pattern Classification, 2nd Ed. (Wiley, 2001). He is a member of the Audio Engineering Society, the Acoustical Society of America, a Fellow of the IEEE, and a Fellow of the American Association for Artificial
Intelligence.

Directions

http://www.dolby.com/company/directions_sf.html