Path: utzoo!attcan!uunet!husc6!bbn!jr@bbn.com
From: jr@bbn.com (John Robinson)
Newsgroups: comp.sys.next
Subject: Re: A sample of Voice Mail!
Message-ID: <32291@bbn.COM>
Date: 15 Nov 88 06:55:15 GMT
References: <7329@orstcs.CS.ORST.EDU> <13401@oberon.USC.EDU> <5284@thorin.cs.unc.edu>
Sender: news@bbn.COM
Reply-To: jr@bbn.com (John Robinson)
Organization: BBN Systems and Technologies Corporation, Cambridge MA
Lines: 18
In-reply-to: steele@unc.cs.unc.edu (Oliver Steele)

In article <5284@thorin.cs.unc.edu>, steele@unc (Oliver Steele) writes:
>Once I trimmed the header and the spaces
>at the beginnings of lines, I got
>	voice		47060
>	voice.Z		32775	(30.4% savings)
>	voice.uu	45177	( 6.0% savings)
>for the original file, the compress-ed file, and the compress+uuencode-ed
>one, which is hardly as dramatic.
... for what started out as 5 seconds of voice, maybe enough for 20
words, or say 200 characters typed.  I think I felt usenet tremble :-}

Soooo... speech understanding is clearly a better compresser than
compress.  Not too surprising; moreover, low-rate speech digitization
systems use models of speech to get that performance (that is, they
reduce my speech to something like phonemes-of-English-spoken-by-me.)
--
/jr
jr@bbn.com or bbn!jr