[aprssig] an off-topic query about statistical analysis indigital modes

[Chris Howard]

On Fri, 21 Sep 2007 09:31:14 +0100
"Dave Baxter" <dave at emv.co.uk> wrote:

> As of course does Morse code!
> 
> Thanks for the book link btw Scott.
> 
> Dave G0WBX.

I knew about low-level things like Huffman encoding.
What I was asking about is a higher level dictionary
of words and phrases and probability rating.

For example, if the string "tje" appears
in a keyboard-typed message it is likely to really
mean "the", or maybe it just might mean "tye" but that
is much less likely.  You can do that at the bit level
and again at a semantic level.

My BlackBerry cellphone/pda has a keypad with
two letters assigned to every key.  Usually I can type
my message just as if the keys were given a single letter
and the BlackBerry algorithm figures out that I meant
"the" not "ygr". So it is effectively a 2 to 1 compression
of the communication between my fingers and the device.

I know just enough about the whole issue to ask dumb
questions.  So I think that book will be very helpful.

Chris Howard
w0ep

>  
> 
> > -----Original Message-----
> > From: Steve Noskowicz [mailto:noskosteve at yahoo.com] 
> > Sent: Friday, September 21, 2007 2:30 AM
> > To: TAPR APRS Mailing List
> > Subject: Re: [aprssig] an off-topic query about statistical 
> > analysis indigital modes
> > 
> > 
> > 
> > Perhaps you know that PSK31 has variable length symbols with 
> > more common ones (letters e, t, a, o, i, n, s))  having shorter codes.
> > 
> > 73, Steve, K9DCI
> > 
> > 
> > --- Scott Miller <scott at opentrac.org> wrote:
> > 
> > > http://www.tapr.org/pub_wdcdat.html
> > > 
> > > Wireless Digital Communications: Design and Theory, by Tom 
> > McDermott, 
> > > N5EG.
> > > 
> > > I'm sure there are plenty of other good texts out there, 
> > but this one 
> > > covers a lot of ground that's of particular interest to hams.
> > > 
> > > Also see the Wikipedia articles for convolutional code, viterbi 
> > > decoder,
> > >   and Hamming distance.  They have some good information 
> > and plenty of 
> > > references.
> > > 
> > > Scott
> > > N1VG
> > > 
> > > w0ep at frii.com wrote:
> > > > 
> > > > Is there a good book to learn about these things?
> > > > 
> > > > 
> > > > On Thu, 20 Sep 2007, Scott Miller wrote:
> > > > 
> > > >> I think what you're describing is basically a
> > > convolutional code.
> > > >>
> > > >> Such things work better when you've got
> > > side-channel information to work
> > > >> from - like if you know one particular bit was
> > > closer to the decision
> > > >> threshold than the others, you know that one's
> > > the most likely to be wrong.
> > > >>
> > > >> In APRS you might be able to recover more data by
> > > knowing that certain
> > > >> fields are always digits, or always alphanumeric,
> > > and using that
> > > >> knowledge to try flipping specific bits to see if
> > > the new message
> > > >> matches the checksum.
> > > >>
> > > >> Of course, you're still better off encoding it
> > > with a proper
> > > >> convolutional code in the first place...
> > > >>
> > > >> Scott
> > > >> N1VG
> > > >>
> > > >> w0ep at frii.com wrote:
> > > >>> We've kind of been on an off-topic swing and I know 
> > this is a good 
> > > >>> place to find people that
> > > know
> > > >>> about digital data transmission so:
> > > >>>
> > > >>> Are there any digital data transmission systems
> > > that
> > > >>> use a dictionary of expected terms and
> > > statistical
> > > >>> analysis that assigns probability to incoming raw data to match 
> > > >>> against the dictionary?
> > > >>>
> > > >>> Sort of like what my BlackBerry does with
> > > SureType mode?
> This mail has been scanned by Palmer Cook Computer Services Limited.  www.palmercook.co.uk
> _______________________________________________
> aprssig mailing list
> aprssig at lists.tapr.org
> https://lists.tapr.org/cgi-bin/mailman/listinfo/aprssig