[Writelog] RE: fullcallsign.wav voice-keying

[llindblom at juno.com]

Yes, decoding vice might be fun but I left the o out of voice.

-- llindblom at juno.com wrote:

And in a year or two we will be taking speech recognition in WriteLog that will decode vice, pop the call sign into the entry window and then send the proper voice message, etc without human intervention.  

Domo Arigoto Mr Roboto
73 W0ETC 

-- Eric Hilding <eric at hilding.com> wrote:

Let me clear up some apparent confusion about how this can work (although
as previously stated, it is NOT for everyone).  It is NOT the monster
tasks some are eluding to.

As discussed with another WriteLog user on the phone yesterday, the future
would be to use highly sophisticated speech recognition type stuff to
eliminate the recording of i-n-d-i-v-i-d-u-a-l callsigns, but far surpass
the use of i-n-d-i-v-i-d-u-a-l letters & numbers in terms of reducing the
"robotic" sound.

Most people do not make over 3,000 QSOs in the SS (SSB) contest.  The ARRL
Scores Database isn't up yet for the 2003 SS (SSB) contest, so let's look
at the realities of the 2002 results.

1. Total logs:  1,693
2. Top Scoring operator # of QSOs:  2,658

My original "ballpark" of fullcallsign.wav files to be recorded for the
first go-around was 3,000.  "Interesting"...and far from 20,000 :-)

We know that many SSB SS operators never submit their scores, and are
Annual "Casual" contesters.  That's okay, and why my idea would take the
following approach (simplified)

1. Put all 2003, 2002 & 2001 SSB SS callsigns from the ARRL Scores
Database (i.e., Annual "Serious" contesters) into a database.

2. Combine/pool Cabrillo logs for these same 3 years from amongst a group
of *serious* contesters and strip out the callsigns to add to the callsign
database (this will pick up a tremendous number of "casusal" op callsigns
which don't show on the ARRL submission records.  I already have the db
set up.

3. Purge all duplicate calls to produce an initial list of callsigns to
"pre-record".  If enough logs have been pooled, the initial pre-recording
of FULLCALLSIGNS should get about 70% used in the SSB SS on the first
pass.

4. EXTREMELY IMPORTANT:  The logging software needs to do the following
DURING the contest:  Any callsign  entered into the QSO window for which a
pre-recorded fullcallsign.wav file does NOT exist simply gets appended
(written) to a .txt file.  The reason for this is efficiency in
identifying doing the next batch of pre-recorded callsigns WHICH WILL
IMPROVE THE EFFICIENCY OF THIS WHOLE THING YEAR AFTER YEAR!!!  This is
what I plan to do, and if many of these individual "hit" lists are
shared/pooled with others, then everyone benefits.  If a fullcallsign.wav
file does NOT exist, then individual alpha/numeric .wav files can be
played.  Only a couple of lines of programming code are necessary to make
this all happen.

5. An alternative to the individual alpha/numeric "robotic" files could be
to have BOLD RED TEXT show right above the callsign that reads "GO
LIVE"...which means the fullcallsign.wav file doesn't exist so "GO LIVE".
Either that, or via a .config setting, have the callsign display in a
different color.  There are various options, but obviously, a slight
delay in operator response.

6. My estimate to pre-record 3,000 fullcallsign.wav files is about two
hours and somewhere between 8 to 10 hours to edit for use.  But for most
it will take longer.  I happen to be a Professional Voice-Over artist and
have a 72 track hi-end digital studio.  For those without similar
equipment, I've previously suggested just going to an inexpensive
voice-over studio or small pro recording studio...the cost amortized over
10 years of future contesting use may be worth it to some.  I did a Syphon
Filter voice-over project for Sony several years ago.  They booked me for
four hours.  We got the entire job done in two hours.
"Focused Concentration" can do wonders in this stuff...just like in
contests :-)

7. One of the keys to making pre-recorded fullcallsign.wav files work is
to frequently monitor the "energy" & "delivery" of what you are doing when
recording in order to keep the best match possible of total final output.

8. My recommendation is to record at 44.1Khz 16bit Stereo as a minimum
(for future digital stuff use...known or unknown at present), and simply
to resample the files to 22K MONO for current SSB SS use.  The sample
fullcallsign.wav demo files I've used are only about 50KB each, but those
were short callsigns.  So conservatively guesstimating here, at even 100KB
for each fullcallsign.wav file, and 100KB for each fullserialnumber.wav
file (do I really need to explain how this works too???)... we are looking
at no more than 200KB per QSO.  If you make 3,000 QSOs in the SSB SS
contest, the maximum hard drive space for the files would be about 600MB.
But keep in mind that only 200KB or less will be used during any ONE QSO,
so we are NOT talking monster overhead here.

9. The ideal is to have ONE "hot-key"/Macro to be able to toggle back &
forth between Automation and "Live".  Automation could be for slow periods
or if one's voice starts to go...any combination of uses desired.  Using
primarily Automation (a/k/a FULL VOICE KEYING), my estimate is that
operator fatigure during the contest can be reduced by at least 30%.

10. This is NOT for everyone.

73...

Rick, K6VVA

_______________________________________________
WriteLog mailing list
WriteLog at contesting.com
http://lists.contesting.com/mailman/listinfo/writelog
WriteLog on the web:  http://www.writelog.com/

_______________________________________________
WriteLog mailing list
WriteLog at contesting.com
http://lists.contesting.com/mailman/listinfo/writelog
WriteLog on the web:  http://www.writelog.com/