00100			Status Report on Project Amanuensis
00101	
00150							December 4, 1973
00175	
00187		The Amanuensis Speech Recognition System, under development
00193	at the Stanford A.I. Laboratory, is a signature-table oriented system
00194	that uses machine learning techniques and attempts to extract the 
00195	maximum amount of linguistic infomation from the acoustic speech
00196	signal. It differs from the system previously described (see attatched
00197	report) in a number of important respects.
00200	
00300	Personnel
00400	
00500	    James L. Hieronymus (post PhD, unsupported but working full time)
00600	    Neil M. Miller      (post PhD, supported full time)
00700	    Bertrand Meyer        (graduate student, unsupported)
00800	    Arthur L. Samuel    (supported 1/4 but working full time)
00900	
01000	Present activities
01100	
01200	    Hieronymus
01300		Just finishing a working paper outlining plans (copy attached).
01400		Debugging stage on his new acoustic segmenting program.
01500		Preparing paper for the April IEEE Symposium on Speech Recognition.
01600		Revising feature assignments for phones.
01650		Studying acoustic coorelates of transitional phonemes and
01675		transitions between phonemes.
01700	
01800	    Miller
01900		Just finished work on his new pitch detector.
02000		Preparing paper for the April IEEE Symposium.
02100		Starting work on a formant tracker which will take advantage of
02200		the improved FFT's that result from pitch synchronous analysis.
02300	
02400	    Meyer
02500		Adapting existing signature table techniques to the problem of
02600		going from a phonetic to a phonemic representation.
02700	
02800	    Samuel
02900		Has just finished translating the entire existing library of
03000		working programs from a fixed interval basis to a pitch
03100		synchronous basis.
03200		Working on improvements in the learning techniques to minimize
03300		the amount of hand labelling of learning material that is
03400		required.
03500		Writing paper for the April IEEE Symposium.
03600