Hardware-Based Speech Recognition?

[Janak Parekh]

<div class='os_post_top_link'><a href='http://www.theregister.co.uk/2004/09/14/speech_silicon/' target='_blank'>http://www.theregister.co.uk/2004/0...speech_silicon/</a><br /><br /></div><i>"The US National Science Foundation has awarded a $1m grant to researchers in the US who want to put speech recognition on a chip, a move the project's proponents claim will revolutionise the way we communicate. Rob Rutenbar, Jatras professor of electrical and computer engineering and computer science at Carnegie Mellon University, will lead the project. The research will be conducted in tandem with scientists from the University of California in Berkeley."</i><br /><br />The article suggests that moving speech recognition algorithms into silicon is key for lower power and computation requirements. The article is highly speculative about the ramifications, but it would certainly be cool if it could become a reality. Would you want speech recognition in your Pocket PC? ;)

[nuka_t]

i say at least 5 years before this tech is pocketsized. it will hti the desktop first, and it will not be a hit there i dont think, but once they shrink it and put it in a pda, im buyin'

[Dolphin]

YES! YES! YES!

BTW... how well does MS Voice Comand work on the PPC?

http://www.microsoft.com/windowsmobile/downloads/voicecommand/features.mspx

[jgalindo]

YES! YES! YES!

BTW... how well does MS Voice Comand work on the PPC?

http://www.microsoft.com/windowsmobile/downloads/voicecommand/features.mspx


I had siemens sx56 and now have ipaq 6315. Msft voice works great.

[Dolphin]

YES! YES! YES!

BTW... how well does MS Voice Comand work on the PPC?

http://www.microsoft.com/windowsmobile/downloads/voicecommand/features.mspx


I had siemens sx56 and now have ipaq 6315. Msft voice works great.

Thanks... I think I'll give it a try!

[mscdex]

There's another one out there too, made by Fonix here (http://www.fonix.com/page.cfm?name=product_voicecentral_for_pocketpc&ID=13)

[Darius Wey]

MS Voice Command works quite well, although if you have a twist to your English accent, it may not work too well. :roll:

Hopefully future versions of Voice Command will accommodate for this.

[kzemach]

I have mixed feelings about this. As probably known by everyone in this forum (and all their close friends) there are already software based translation tools on the PPC, be they the Darpa-funded Phraselator or some of the other commercial ones that I can't remember off the top of my head. Granted, while the Phraselator stuff is pretty rough and is marginal at best in the ability to do the task, if there's one thing we've seen over the years, it's to never underestimate the rapid pace of hardware speed and energy efficiency increases.

I've picked on the Phraselator a lot in a few speech forums here, but while I don't think it really works right now, it seems pretty clear that at some point in the future it WILL work.

Not saying I'm against specific hardware development; there are certainly plenty of aftermarket video cards for both display and capture, because the build in stuff barely, if at all, works for those applications. However, in a day and age where most of the Darpa-esque/military S&T funding and is to move stuff FROM hardware to software based (e.g. software radios), it is kind of curious that they'd be doing this. But I guess we won't know until we try, and I don't personally know enough about the back end of speech recognition to second guess the researchers.

I guess my gut seems to say that the hardware they'll develop in the proposed 2-3 year timeframe will be lagging behind the software-based work in 2-3 years coupled with the impending improvements in basic chip speeds and power efficiencies. Then again, my gut isn't exactly a speech recognition expert.

[surur]

They would do better to improve the basics of speech recognition e.g DIgital Signal Processing and Single Instruction, Multiple Data functionality to mobile chips, instead of trying to fit a whole recognition model on a chip.

I dont think in any case people like talking to their gadgets, else speech recognition would have taken off on the desktop, where there is no shortage of power. It would be even worse on the subway, talking to your PDA.

The real scandal of PDA's are that they do not have any floating point processor, which is a real deficit in this day and age.

Surur

[humayunl]

While we're on the voice recognition topic, I am thinking about installing a voice recognition app. on my PPC (XDA II) and want to know what options I have and how good they are.

So far I know of Fonix Voice Dial and there's MSFT Voice Command.

Are there any more options? obviously the speed and accuracy of recognition is what is key here but besides that is there anything else I should be looking for? do these apps let you control all functions of the PPC or just the phone/dialing part etc.?

Any feedback from experienced voice users would be appreciated.

[whydidnt]

Fonix has a whole suite of Voice related apps, form dialing to reading your e-mail to you, etc.

However great voice recognition sounds, I'm still trying to figure how it could really be useful. Think about how you input information today. Typically somebody tells you something, and you enter it on your PPC, using a stylus or keyboard. Now think about that same situation, only instead of typing it, you speak it to your PPC. Don't you think the person providing the information is going to be little uncomfortable listening to you paraphrase what they said. Then take it a step further. If we have a whole room full of devices such as this everyone will be talking into their devices at once?

Don't misunderstand, I love the concept, but need help understanding how it can be used in a non-awkward way publically.

[Bruce Babcock]

Don't misunderstand, I love the concept, but need help understanding how it can be used in a non-awkward way publically.

Why isn't the voice app transcribing conversation/meeting for you!
It would be a lot less disruptive than trying to write/type and pay attention to what is going on.
:wink:

[Sven Johannsen]

I'm thinking similarly to whydidnt. The social ramafications of speech enabled personal electronics is mind boggling, and not in a good way, IMHO. There have already been numerous discussions about the advent of cell phones being allowed on airplanes and the resulting cacophony of folks shouting into their Nokia's. Of course you know you have to talk louder to use a phone than normal conversation. :roll: I am going to assume this will be true of PDAs as well. If it doesn't get you the first time, talk louder. :?

While there are certain situations that speech to text, or voice command and the associated text to speech would be priceless, I'm afraid it would get very annoying, very fast, since I don't see individual courtesy getting any better than it is today with cell phone use.

[humayunl]

I agree that in public, the use of voice activation could be disruptive, intrusive or just plain confusing (given enough people doing it in one place), but that only means that there are certain situations where perhaps an alternative input menthod is better suited.

I have never used voice activated anyting, and am just going to try out the Fonix of MS app. so I cant speak from experience, but from what I can imagine, i probably would not use it in public.. I would start with seeing whether it really makes it convenient for me to place calls and lookup contacts, locate and open programs and files. If it does all that fast and accurately enough, I'll be happy.

But i wouldnt be doing this in public, if for no other reason than just because it might make me look silly talking to myself :)

[mcsouth]

But i wouldnt be doing this in public, if for no other reason than just because it might make me look silly talking to myself :)

Every now and then, I actually notice people staring at me while I'm talking on my cellphone in the grocery store, in the mall, etc - not because I'm on the phone, but because I'm not actually holding a phone to my ear. The moment I climb in my car, I plug the hands-free earpiece into my phone, and I find it handy when I'm trying to get groceries, etc. However, it is clear from the stares that most folks find that strange. It is interesting that you don't get that response if you actually have one hand holding the handset to your head!

Consider Europe, where Bluetooth headsets are big business - you don't even have the wire to clue you in, and some of those BT headsets are tiny - they almost fit into your ear! I expect that they are more used to seeing folks "talking to themselves".....at some point, that won't be considered strange anymore.

[Damion Chaplin]

I would only find speech recognition useful if there were text-to-speech as well.
For example, if my iPaq is sitting on the table across the room, I might be able to say 'Mom-phone number' and have it read out the number to me. Or the classic Star Trek one that I always wish I had: [step out of shower] "Computer: Time" "It is fourteen-hundred eleven".
Basically, if I'm close enough to the unit that I can look it up in ye olde fashioned way, I probably won't use speech recognition.

[nuka_t]

Fonix has a whole suite of Voice related apps, form dialing to reading your e-mail to you, etc.

However great voice recognition sounds, I'm still trying to figure how it could really be useful. Think about how you input information today. Typically somebody tells you something, and you enter it on your PPC, using a stylus or keyboard. Now think about that same situation, only instead of typing it, you speak it to your PPC. Don't you think the person providing the information is going to be little uncomfortable listening to you paraphrase what they said. Then take it a step further. If we have a whole room full of devices such as this everyone will be talking into their devices at once?

Don't misunderstand, I love the concept, but need help understanding how it can be used in a non-awkward way publically.

motorolla(i think it was) is making headset type microphones that you can put near your mouth and basically whisper into, without anyone hearing. they are designed for cellphones, but i think that those, coupled with this technology will be the future. face it, keyboards suck.

[jsalsman]

Some speech applications are a long way off, like dictation (http://www.pocketpcthoughts.com/forums/viewtopic.php?t=32140).

One step up from simple command & control is pronunciation diagnosis (http://www.readsay.com/pro.htm), but most speech recognition SDK's don't support that very well.

[ebadger]

Command and control speech recognition can be very useful on mobile devices. I have 675 contacts in my address book. To find a contact using the tactile and gui interface is a chore. I already know the contact that I am looking for in my list of contacts. Why should I have to slog through a visual list to find the contact that I want to access when I can simply speak the name of the contact and bring them up reliably and predictably. I think this case becomes even more compelling while driving in the car. However, I am so used to this method of accessing my contacts now that I use speech interface to dial always.

[Sven Johannsen]

I have 675 contacts in my address book. To find a contact using the tactile and gui interface is a chore.

You might want to take a look at Pocket Contacts, which is an add-in to Pocket Breeze (www.sbsh.net) The contact retrieval methodology is outstanding. It's sort of T9 thing where you start spelling the name and it narrows the options down rapidly.

[Jonthepom]

For deaf people the mobile phone has been a leap in technology. They can now carry on a phone conversation by text. Wouldn't it be nice to have your PPC phone translate a face to face conversation as well.

I would buy it even if it cost $k's.

[jsalsman]

Wouldn't it be nice to have your PPC phone translate a face to face conversation as well. I would buy it even if it cost $k's.

You want Ectaco (http://www.ectaco.com/) "Pocket VT" or Pocket Voice Translator.

Soon thereafter, you may have an epiphany and then want my product (http://www.readsay.com/pro.htm)