00100	COMMENT ⊗   VALID 00003 PAGES
00200	C REC  PAGE   DESCRIPTION
00300	C00001 00001
00400	C00002 00002	
00500	C00005 00003	Proposed Research-
00600	C00007 ENDMK
00700	C⊗;
     

00100			NSF Pre-Proposal
00200	
00300		Computer Simulation of Natural Auditory Signals
00400			and Reverberant Spaces
00500	
00600	
00700	In recent years there have  been  significant  advances  in  computer
00800	simulation  and  analysis  of  complex  auditory signals.  On the one
00900	hand, the use of the  computer  has  demonstrated  the  extraordinary
01000	complexity   of   natural   sounds   and   consequent  difficulty  in
01100	understanding, and on the other, the computer has provided the  means
01200	of implementation of powerful simmulation and analytical techniques.
01300	
01400	At  the  Stanford  Artificial  Intelligence  Laboratory  programs and
01500	techniques have been developed over a period of ten  years,  for  the
01600	simulation  of moving sound sources in reverberant spaces and for the
01700	analysis  and  synthesis  of  complex  signals,  some  of  which  are
01800	startlingly  simple  in  implementation and novel in conception.  The
01900	progress of this research has certainly been  enhanced  by  the  high
02000	level    of   technological   expertise   which   characterizes   the
02100	inter-disciplinary  environment  of   the   Artificial   Intelligence
02200	Labotratory.
02300	
02400	Audio  signals  are  generated  by  a  PDP-10  computer  which  has a
02500	4-channel  D-A  converter  for  analog  conversion  of  the   sampled
02600	waveform.  The  analysis is accomplished by means of an A-D converter
02700	which converts audio input into a sampled wave which is stored  on  a
02800	disk  for processing. The programs forthe synthesis of wave-forms are
02900	based on the acoustical compiler developed by M.V.  Mathews  at  Bell
03000	Telephone  Laboratories  adapted  to  the  time-sharing,  interactive
03100	capabilities of the PDP-10 system, while the wave  analysis  programs
03200	were written by J.A. Moorer of Stanford. Proposed Research-
03300	
03400	1.   Analysis and Synthesis of Music-Instrument Tones and the Singing
03500	Voice-
03600	
03700	The time-variant properties of natural sound in  both  the  frequency
03800	and  amplitude  domaines have been shown to be of critical importance
03900	in the perception of timbre or tone-color.  Programs  and  techniques
04000	are  being  developed  for  the  analysis and synthesis of acoustical
04100	signals which represent these time-variant properties in the form  of
04200	spectral  plots.  The frequency and amplitude data is then reduced to
04300	its  minimum  without  disturbing  the  original  perceptual   image.
04400	Subjective  judgements  are  made  by  musically  trained  listeners.
04500	Manipulation   of   the   simplified   data   in   conjunction   with
04600	multi-dimensiional  scaling  techniques  of  judgements, suggests the
04700	critical subjective cues for the various categories of timbre.
04800	
04900	2.   The  Synthesis  of  Complex  Spectra  by  Means   of   Frequency
05000	Modulation-
05100	
05200	
05300