voice recognition in flash?

[mixedtrigeno]

the subject basically says what i am thinking and it is possible? is there a way to implement voice recognition in flash even if it is by using a third party program and implemented so that it works from within flash?
basically by using the computer mic.

or is this impossible?

[VENGEANCE MX]

If it is possible, it would still be _EXTREMELY_ hard. I doubt even the best coders here could implement it.

[hooligan2001]

I dont know, i thought flash could only convert frequencys. If you could store them somehow in an array. Which would be a huge array full of numbers. Then just run a check on that array every time. Would be pretty hard to achieve the exact same combination of numbers everytime though.

[webgeek]

I don't believe you can do it in Flash. You need access to the raw microphone data. OR you simply pass the raw data to something else that has some power behind it. In that case, you MIGHT be able to pull it off with plugins in FlashCom. At that point though, you arn't using Flash anymore. This is a good example of when to just move on from Flash to something else

[RipX]

Could be done using a neural network...i've been working on one of these on and off for a while, but it would have limited results due to the itteration speed using flash and the number of frequencies required!

RipX

[ihoss.com]

not using only flash. Flash can use the microphone (and the webcam) but can only recognise changes in volume, not frequency. Same with the webcam. It can detect motion but not recognise images or even use the images. It can, however, send input from the mic and webcam over a network using another macromedia tool (I can't remember the name).

[mixedtrigeno]

Originally posted by RipX
Could be done using a neural network...i've been working on one of these on and off for a while, but it would have limited results due to the itteration speed using flash and the number of frequencies required!

RipX

so it can basically be done using another software along with flash...just basic one word. not no sentences and the words are pre-assign to flash. for example:

code:

---

array=[[stay,fetch,walk,etc]];



if(you say stay!){

//check array for words

//maybe with a for loop to check for the words within the array//

//then something here to communicate with the other software and microphone//

if(stay==true){

_root.dog.gotoAndStop("stay");

}

---

thats the idea

[webgeek]

>>Could be done using a neural network...i've been working on one of these on and off for a while, but it would have limited results due to the itteration speed using flash and the number of frequencies required!

Not in Flash directly. You don't have access to frequency data, just volume. You are not going to get even minimally accurate results without frequency information. You would need to use a NN in FlashCom to do it, assuming FC has access to frequency.

[mixedtrigeno]

so basically you would need a software that recognizes frequencies that can work with flash?

[webgeek]

Ok, time to move beyond Flash here

Flash, as development/technology platforms go, is VERY weak. You are tightly coupled to the APIs that Macromedia provides and have no way to extend them. The BEST you can do with Flash directly is call out to other applications and platforms using a variety of integration techniques and ask them to do work for you ie. XMLSocket, XML.send/sendAndLoad/load, LoadVars, FlashCom, and various others. Even high-ish level languages like Java, PERL, PHP, C#, etc. allow you to step out of their sandbox and work with low-level languages like C and C++ using various techniques. C and C++ can do the same thing down to assembler. The point is that all of these languages give you more power by letting you go a lower level if you need to. Flash doesn't. This is where Flash limits you.

When you ask to do something unique in Flash, you have to first see if any of the built-in APIs provide it. Can I record the data with Flash and then interpret it? No, not really. I CAN take the data in real-time though. Lets assume thats good enough. Assuming that voice recognition requires an understanding of the audio stream's changing frequency we run into the next stumbling block. Flash can't give you info about the frequency. The best you can do with Flash is volume. This is pretty good, but not what you need.

Next you need to see if Flash provides a way for you to get the data out of Flash and into an area where you can work with it. Do any of Flash's integration capabilities let me send audio to someone else? Yes, just one, FlashCom can receive an audio stream from Flash. FlashCom provides some lower-level capabilities and gives you one excellent ability; you can use Java inside of it (I believe so anyways).

In conclusion, you MIGHT (I've never used FlashCom to this degree, so I'm speculating a bit) be able to use Flash to get an audio stream from a user's microphone, stream it to Flash Com, which uses Java code to invoke a voice recognition library or API that can parse the voice into text. Then the code uses FlashCom APIs to send that text data back to Flash.

What a pain in the butt. And what do you get for all that effort? A barely functioning, extremely slow, very inaccurate, and expensive voice recognition system. If this is an academic exercise that you want to do just to see if you can, well then more power to you and good luck. If this is for a real project, just leave Flash behind and jump to a language that can do it directly. Microsoft and IBM both have excellent speach recognition APIs you can develop against.

Use the right tool for the right job. Just because theres a hammer in your hand doesn't mean everything is a nail.

[RipX]

yes... you'd need to get a seperate program to create an array of frequencies then use the neural network to learn the patterns of specific words spoken by various people!

RipX

[Beatcow]

Lol, I was just working on one of these last week in flash. I figured that if I toggle the mic to be sensitive enough, the volume that the mic receives every frame would be graphed out and later on compared to other, prerecorded words.

.

[mixedtrigeno]

thanks a million webgeek your information is extremely helpful. and yes it is a personal project. and by the information you provided me shows that it is extremely hard to accomplish this. however going by wat ripx is saying it could be done. i know for a fact i wont be able to do this for the simple reason that this is way too much for me.

ripx if you are ever able to accomplish this extremely hard task can you please pm me. if it is not too much to ask.

and now in case you guys are wandering why i asked about this is because i wanted to see if i can do a RAINBOW 6 TYPE GAME where you control a group of soldiers and you click on one and then you tell him from a list of commands like (hold,fire,move-in,snipe,follow,etc)this are pre-asign words and then the soldier that you have clicked on performs the action. if you played RAINBOW 6 or SOCOM you know what i mean. well that is the reason why i was asking about it but if the voice recognition gets too hard ill do it with typing.
the same idea but with typing.hmmmm hmmmm it will be a hundred times easier...RIGHT? at least in theory and the way i am thinking but if someone can get the voice recognition done and is willing enough to share with me i guess i can stick to that idea.