More About Speech Recognition for the iPAQ

[EmXtrix]

An article in the New York Times Technology section talks about upcoming speech recognition products for the iPAQ (among others).<p><font face=verdana size=2><b>Read More:</b> <a href="http://www.pocketnow.com/cgi-bin/directory/detailed.cgi?db=news&id=461">More About Speech Recognition for the iPAQ</a></p>

[Unregistered]

Am I missing something? My Jornada 568 came with voice recognition software which doesn't require "training" and works with anyone's voice. It also talks back to prompt or confirm commands. Works very well.

[EmXtrix]

I use the Jornada speech recognition too, but it isn't really "speech recognition," it is more like "voice navigation." I think the new iPAQ PPCs might have something else... I can't wait to find out.

[Russ Smith]

The Jornada has what I refered to as voice command. It's a subset of voice recognition that allows simple control. The IBM program uses natural language input. (You'd say "How are you feeling?" and your Pocket PC would respond "My main battery is at 68% charge and my backup battery is good." or something like that.) At a level lower than that is voice input or voice dictation, which allows you to speak to your Pocket PC and have it appear as though you typed it.

Voice command is pretty close to perfected technology. Phone companies use it to navigate their systems and some phones use it to dial hands-free. The system doesn't actually care what the sound it hears is. It just associates it with a command. (Theoretically, you could allow your cat to page down for you.) Dictation is harder and involves parsing sentances and recognizing spelling based on the surrounding words. (Cognates like "there", "their", and "they're" make life difficult.) Natural language input is even harder as it involves "inferring" what you mean when you ask a question. You should be able to ask it for something in any way which makes sense to you. Right now, with each level of complexity, the need to train it to your voice becomes more necessary.

[Frostberg]

Wow I cant wait until the day when I can just ditcate a memo to my PPC and it types it out pretty clost to 100% accurate...but thats gonna be awile and will take up a LOT of RAM. I Bought IBM's Via Voice Gold for $99 when it first came out in 97 or 98 and to be honest, it sucked. It was amazingly off. I did all of the tutorials so it would "learn" your voice and it still didnt help. If you said "Billy went to the park" it would type in "Johnny sailed after to the beach" It was like the program was drunk or something. It was pretty bad and prolly the worst hundred bucks i have spent. It also was bad because in between words you had to pause for like 2 seconds, so you could type faster than the drunken Via Voice. But in just a few years things dramatically improved when the newest Dragon Naturally Speaking came out and I must say it is impressive. You talk at a normal pace, and the recignition is about 95%. So I think within a few years we will be able to use something like Naturally Speaking and dictitate to our PPC, but like I said it is going to be very large to know basically a dictionary of words, and then to learn your voice. Naturally Speaking is 300MBs or so, but in a few years you will prolly have a PDA with 256MBs of onboard RAM and 1GB SD cards or something like that

[Jonathan]

What should be really interesting is if IBM's software will run on the Jornada. I know for a fact that the Jornada's Mobile Conversay does run on the iPaq. I installed it about a week ago. Its runs just as well as on the Jornada.