Keyword scanning/speech recognition

[pgut001@cs.auckland.ac.nz]

I was talking to someone recently about the feasibilty of keyword-scanning
phone conversations.  He thought it was probably still beyond the reach of
current technology, I thought it wasn't (I gave a couple of references in a
recent paper on government attitudes to crypto which indicate that it's being
used right now by a number of governments).  Anyway, I've got bits and pieces
of one or two papers here which people might find interesting.  The first one
is:
 
  "Digital Circuit Techniques for Speech Analysis" by G.L.Clapper, presented at
   the AIEE Winter General Meeting in January 1962.
 
I've only got the first two pages of the paper here, I think the full thing
might have been published in the IEEE Trans.Communications in about 1963.  This
paper mentions a "digit recognizer" built at Bell Labs in 1952, and a Japanese
voice-operated typewriter using 3,000 transistors and 6,000 diodes.  The paper
goes on to describe a means of producing a "compact digital code expressing
significant qualities of speech in a form suitable for machine utilization".
This was in 1962!
 
A more recent paper is:
 
  "Discrete Utterance Speech Recognition without Time Alignment", John Shore
   and David Burton, IEEE Trans.Information Theory, Vol.29, No.4 (July 1983),
   p.473.
 
This generates a feature vector every 10-30ms from input speech which is
compared to pre-generated reference sequences.  It also has references to many
other papers covering the same area.
 
Certainly it appears that by the early 1980's it was possible to scan speech
for keywords and/or speakers and use this to target the surrounding
conversation, and that early work on this had been done since about 1960.  This
is all from public sources, since certain organisations have a *far* greater
interest in this particular area than anyone performing public research it's
likely that equivalent classified research was some way ahead of this.
 
Peter.