Is there any possibility that we can control our N900 through voice commands,can we train it to do things through voice recoginition for example we speak and the N900 writes sms itself. we say stop to stop the music. Infact everything what we do through buttons and touch screen should be done through a voice recoginition software. Can it be developed?

Edit by Flandry: This has been merged into the brainstorm proposal thread. The proposal is here: http://maemo.org/community/brainstor...o_audio_ui-aui

No One intereseted except for me. I guess so...

I would be interesting in something like this. Would be a nice feature to have, even if its only for a few commands.

20 years ago got telephone with calling party
voice selection.
So voice recognition technology is easy, but should be supported by voice recognition chip, hardware solution, not another 1,000+ libraries to slow down a smart phone.

And pray tell...what "voice recognition chip" are you referring to?

Sometimes, you _REALLY_ ... _REALLY_ need to shut up.

You show your complete lack of knowledge, and abundance of opinion, where honestly, it adds nothing to the conversation.

-Thom

CMU Sphinx can be used to build a limited domain voice recognition system that can run nicely within the confines of the phone.

Either Sphinx II or PocketSphinx (with its fast integer transforms), can be used for this.

1. Use CMU Sphinx
2. Create a dbus bridge.
3. use these two bits to provide an API to expose to the rest of the stack.
4. Build appropriate UI using the dbus calls.

done.

-Thom

Can you help me promoting this development.
How to procedd with brainstrom?
how to put this idea in developers' minds.


Regards

Real useful voice recognition is NOT easy. Sure, recognizing the difference between "yes" and "no" is a joke, but if you want to dictate to a phone (like you can with android), its a whole other ball game.

But i still think it will be a very useful tool when ur hands are not free. e.g. while driving, etc

again... I wish you guys would stop spouting off comments that add nothing to the conversation.

I have extensive experience in this field.

Sphinx II, and PocketSphinx can be used for the limited domain aspects of speech recognition for command actuation. The hard parts are:

(1) building a language model with the bandwidth characteristics of the phone. I believe that CMU has released their phone model that can be used as a baseline for this. (they are using it in their Communicator project)

(2) building an agreed upon actuation modality (how to activate voice actuation) and from that modality build a consistent syntax and grammar.

(3) building a corpus against the language model for the agreed upon (2)

With sphinx, limited domain recognition can be done close enough to approximate real time use.

With sphinx's greater domain dictation recognizer, significantly more processing time is needed, and can't be done in real time.

-Thom