|
|
 | | From: | James Salsman | | Subject: | Re: dictation from cellphones? | | Date: | Fri, 17 Dec 2004 11:59:01 GMT |
|
|
 | >>Frankly, with MMS, and the bandwidth that some of these phones
>> operate at when they are in data-mode, only in poor-signal
>> areas would there be a problem.
>
> Are you speculating or is this something you have tested?
I have tested MMS phones. The AMR audio was apparently recorded
at the higher quality AMR rate by default.
> Speech recognition software does not tolerate minor imperfections
> in sound input.
That depends on the kind of imperfection. If there's five harmonics
of 60 Hz cutting a set of notches into the spectrum as a hum or a
buzz, then that's bad. If there's a VOX transmission delay that
makes it sound like you said, "step for man," when you actually said,
"step for a man," then that's bad too.
But, if your audio frames are check-summed and retransmitted if
corrupt, as happens after you "send" a MMS message, then the
recording is going to be as good as or better than you heard it.
I think the big question is what comes next. If the recognition
is done on a server, then the results can be presented in the
form of a textual email MMS reply, but I don't know the best way
to have a cellphone proofreader select alternates for mistakes,
because if that isn't done in real time, interactively with
recognition, then the whole rest of the recognition can be thrown
off-track. I suppose the dictation system could send back a
sentence at a time, and a null reply (or a reply with the sentence
as sent) would mean, "correct; proceed to next sentence," and
there could be all kinds of short non-word mnemonic codes. E.g.,
Message:
It's hard to wreck a nice beach.
Reply:
> It's hard to wreck
..no recognize .cont
[or:] .no recognize speech. .next
It all depends how much people can easily type. Some handsets
have much easier typewriter-style keyboards than others. I can't
imaging what it would be like to send one email for each sentence.
There would have to be a way to do multiple sentences at a time
when they are short.
Sincerely,
James
--
www.readsay.com - maker of the ReadSay PROnounce English literacy system
400 MHz PDA included: $499 -- http://www.readsay.com/PROnounce.html
|
|
|
