OSDev.org

I had an Idea for a new os. Tell me what you think....

It will include speech recognition and gesture recognition.
It will include optional keyboard input but not mouse.
it will start a revolution in the computer industry.
(Think of it, Touchless Computing)


Hmm..............................................................................

And how would you be able to create one if non of these techniques are realtime available to you?! Fantasizing is OK, but do it in your mind not on these forums, please!

Didn't mean to "Disrupt" you... but gesture and speech recognition is very real.... (just google it)

What I mean is, that this is not available to you yet, I asume you are not a millionair or something, cause getting your hands on such equipment will cost you lots en lots of money, which you probably do not have

What "such equipment" are you talking about?

His idea seems perfectly feasable to me.

The ONLY "Such equipment" is a basic microphone and a basic web cam.

There are NO multimillion dollar devices required. The only thing hard to obtain will be speech recognition and synthesis librarys.

And that is the equipment I meant

btw, here in holland we already have such machines, i.e. economical refugees who come here have to do a test if they understand holland Anyway, this is done by speech recognition, also I believe the Japanese are very far with gesture recognition, so nothing new to invent there. Also you'll have to support lots of hardware since you didn't say it will only run on specified hardware.

Just my 2 cents.

I think its a good idea but not for a starting osdever I could not even imagine having to build it....



also the thing about trying not to have keyboards and mice
think abuot in a big office how loud that would be and have to have a seperator wall between EVERYONE or it might pickup someone elses gesture/voice

also I type much faster than speek but thats just my 2.52944 cents and a half

Voice recogition is everywhere now-a-days. My anchient cellphone will dial whatever number I speak into it. It's a big goal, I agree, but I don't think it's that unreasonable for a competant coder.

Do you have any background in audio synthesis and/or processsing jakelstr? The basics are key, here. If you can't currently perform a simple FFT on a soundwave, then I wouldn't even think about trying to implement this... thankfully, however, this kind of audio processing background can be "easily" obtained through the internet. You'll need to invest a lot of R&D into this aspect of the OS, I think (as with all others, of course ) but I think it's a noble goal

As for mouse gestures... these are easy. Take your mouse path, and transform it into a set a vectors, and compare it against known vectors... you can find code and tutorials on this online.

Again, I agree these are fairly difficult goals... but doable.

--Jeff

*edit* As a side-note, my audio processing background is not where I'd like to be, myself. My intention in my second paragraph was not to belittle, but to suggest that a *very* deep audio background will be required -- more then most have, I think. Just wanted to clarify that, as I think some might take the words too heavily (if such a word exists...)

What's actually so OS specific about the idea ? To me it rather sound like a new input concept that could aswell be used with traditional systems (similiar to a graphics tablet). If you just wrote a windows-driver that implements your ideas you could probably make millions. Now just ask yourself why nobody has done this by now



But it makes a huge difference if you only have to recognize a limited set of words/phrases ("one" - "two" - "dial") or the huge amount of program names and paths that might be used on a computer. Your speech recognition would have to be intelligent enough to translate whatever I say into a pain ASCII string that can then be run the traditional way. It's certainly possible to achieve this, but don't expect it to be trivial..

Apart from that I somehow failt to see how it's any more comfortable to say a command rather than just writing it. To me speech-recognition only make sense if it allows me to express myself in my natural language. That basically means that I would have to be able to really talk to my computer, just like Captain Picard does: "Open pictureX from my desktop, resize it to 1024*768 and then save the changes". For this the computer would however really have to understand what I mean. It will probably take many decades until the develeopments in AI have reached a level where such things become feasible..



I think he was actually talking about the more analogous gestures . Just like Tom Cruise in "Minority Report", who uses some special gloves to move windows on his glas screen (gloves, screen). It's probably the same idea as with mouse gestures, but way more complex and expensive (you do need special hardware)..

regards,
gaf

I think you could do the tom cruise idea, very easy by using a web cam, like you can get the games for ps2 (eye toy) http://www.game.co.uk/ViewProduct.aspx? ... mid=323251
But you could move windows instead, a touch screen with out touching the screen .

PS: This project looks interesting http://fre3space.dforge.cse.ucsc.edu/

Just in case that I'm not the only one who really hasn't the slightest idea about game consoles: You can find a video showing some Bruce Lee wannabe in front of his eye-toy on youtube.

The impression that I got is that eye-toy doesn't really recognize anything, but only reacts to motion. As the background is static the only thing moving is the player. It's therefore quite trivial to detect his motions by just comparing the last few frames recorded by the camera (not per pixel - you get the idea). The only game that might be a tad more sophisticated is antigrav, although I guess that the basic idea remains the same.

While such an approach might be fine for all kinds of "visual karaoke" games, it's certainly not sufficient for a desktop system. After all I'm probably not the only one who doesn't want to do 2 kick-flips just to close a window . What I actually had in mind is a system as it might be used by divers or soldiers during urban warfare. In both situations you can't talk and thus have to try to express yourself by small gestures of your hands.

I could for example imagine to just point to a window to activate it. As long as I keep my hand pointing at it the window follows my finger. If I want to close a window I activate it by pointing to it and then just snap with my fingers (that should look really cool..).

As you can guess, such a system is much harder to implement as it requires the computer to really recognize what's going on. Just detecting motions and reacting to them in a mechanical way isn't enought. I would also expect, that such a system can only be implemented using hardware accelertion, as it would otherwise eat up too half of the CPU time..


These guys just took an radio emitter and triangulated it using three different receivers. This merely replaces a mouse and doesn't allow any real guestures. I also don't see how it's any more comfortable to stand in front of the screen waving your mouse through the air: Call me old fashioned, but I like my beer cold, my TV loud and my mouse on its pad

regards,
gaf

GO AHED man,
try to make it work rather than trying to spend time
arguing...Actions speak lauder than words