First, remember what you have taught, you need to formalize the requirements of the project first to start it. What exactly this plugin is going to implement, what kind of speech recognition:

 * Single person or multiple ones

 * Recognize commands or dictate

 * If commands which commands, what will the do exactly.

Do you have this written down already?

--- (Edited on 1/18/2010 12:36 [GMT+0300] by nsh) ---

I thought to use

-single person

-recognize commands

-commands are, when person says "connect google", it should connect to gogle, "back", "go", "exit"...commands also work like that

--- (Edited on 1/19/2010 1:20 am [GMT-0600] by Visitor) ---

Im really thakful to you if you can help me

--- (Edited on 1/19/2010 1:25 am [GMT-0600] by Visitor) ---

>create a speech recognition plugin to the firefox. im doing this as my

>research project.

Why a FF plugin?  Look at Ivela-web's approach to speech recognition on the web:

[...] when you load the page, my applet will search for the lastest version of julius under c:\julius (hidden folder in windows) or <your_home>/.julius (linux case). If there is no julian neither its grammars for our tests in you Hard Drive, my applet will download the suitable zip archive for each operational system.

After the download, the zip archive is unpacked and julian is ready to run. This happens only for the first time you load the page. In conclusion, it is a FAT client running julius. That's why th applet is signed.

The same could be done with Sphinx.

Ken

--- (Edited on 1/19/2010 1:45 pm [GMT-0500] by kmaclean) ---

He probably wants a recognition to happen on each page, not in a single page, doesn't he?

--- (Edited on 1/20/2010 01:53 [GMT+0300] by nsh) ---

Well, next step is to read documentation and create a simple plugin that will do nothing except starting. Documentation for that could be found on

https://developer.mozilla.org/en/Plugins

Once you'll have one it's possible to add speech recognition functionality to your plugin.

--- (Edited on 1/20/2010 01:57 [GMT+0300] by nsh) ---

>He probably wants a recognition to happen on each page, not in a single

>page, doesn't he?

Good point!

--- (Edited on 1/19/2010 7:33 pm [GMT-0500] by kmaclean) ---

rdus, let us know how you get along with this.

I recently looked at a similar scheme. It is easy enough to write an extension for Firefox which would open up a helper window to receive verbal commands and fire them at the parent window and get it to do various things like take the focus, or display a message box.

However you soon run into an issue called "cross site scripting" which pretty much stops the interesting stuff. For security reasons some methods are denied to pages that are not in the same domain. For example, I found that attempting to send a "scrollTo(x,x)" javascript command (which appears innocent enough!) gets denied with a permissions failure.

--- (Edited on 2/2/2010 12:37 pm [GMT-0600] by colbec) ---

Check out something like this:

http://www.speechapi.com/

--- (Edited on 2/3/2010 5:50 pm [GMT-0600] by pomprocker) ---