Newsgroups: comp.speech,sci.electronics,comp.arch.embedded
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.sprintlink.net!news.indirect.com!seeker
From: seeker@indirect.com (Stan Eker)
Subject: Re: Speech for embedded systems
Message-ID: <D9Dt7J.BH8@indirect.com>
Sender: usenet@indirect.com (Internet Direct Admin)
Organization: Internet Direct, indirect.com
Date: Tue, 30 May 1995 08:06:54 GMT
References: <D9CFz7.498@magi.com>
X-Newsreader: TIN [version 1.2 PL2]
Followup-To: comp.speech,sci.electronics,comp.arch.embedded
Lines: 40
Xref: glinda.oz.cs.cmu.edu comp.speech:5746 sci.electronics:134467 comp.arch.embedded:1264

Richard Sloan (etm@magi.com) pondered:
: I am wondering if anyone can give me any suggestions for 
: hardware that will allow pre recorded speech to be 
: played on a small microcontroller / embedded system. 
:  
: I am looking for a system with minimal part count, ease of 
: recording speech like through a sound blaster card, low cost 
: and easy interface to 68HC11/8031 type controllers, speech data 
: stored in eprom. 
:  
: I would also like to extend this idea to being able to record 
: right in the embedded system into static ram, but this is not 
: as important as the playback. 

Adding a CODEC to an 'HC11 or 8031 is only a couple of chips and a small
handful of passive parts, but the STORAGE is a bitch.  There's a bunch of
different methods of compressing speech data down to reduce bandwidth and
storage requirements, and the overall system cost would probably benefit if
you didn't have to access hundreds of Kbytes or megs of SRAM/EPROM/EEPROM.

You might check with the folks over on comp.dsp to see what's the current
favorite for speech compression, but be nice and look at the FAQ (found on
rtfm.mit.edu in /pub/usenet-by-group/comp.dsp/comp.dsp_FAQ*) before you
annoy 'em with something that's been asked frequently.  A quick browse of
the old copy I have is filled with info on it, and even lists some existing
hardware that'll allow you to tack up to a couple of HOURS of speech onto
your XYZ project.

Simple math: sample 1 channel at 8KHz (minimum for useable speech), 8-bit,
no compression.  How long does it take to completely fill the normal memory
space (64K) of an 'HC11 or 8031?  Answer: 8 seconds.  With some added
decoding you can expand that time, but you'll find that 1M SRAM chips (128K
by 8-bits) are pricey, and the 4M SRAM chips are much worse.  Without some
sort of compression, you're spending most of your money and space on memory.

You might also check to see if comp.speech and comp.compression have FAQs.
If you just blindly post to all 3 simultaneously, they'll probably ignore
you.  Get out of the habit, OK?  It's rude and obnoxious.

  
