Active Topics

 



Notices


Reply
Thread Tools
ZogG's Avatar
Posts: 1,389 | Thanked: 1,857 times | Joined on Feb 2010 @ Israel
#11
Originally Posted by taixzo View Post
Unfortunately I do not have an N9, and I don't have the money to get one. If someone else is interested in getting it running though, it should be simoke in theory. Is pocketsphinx in the N9 repos?
there is some package sphinx and log of #harmattan @ freenode gave this :
javispedro npm, btw, I write a Maliit input plugin to add dictation support to all harmattan apps using pocketsphinx
javispedro npm, https://gitorious.org/javispedro-n950-misc/maesphinx
 
ZogG's Avatar
Posts: 1,389 | Thanked: 1,857 times | Joined on Feb 2010 @ Israel
#12
here we go, i think it's the same package — http://harmattan-dev.nokia.com/pool/...free/s/sphinx/
 
Posts: 1,086 | Thanked: 2,964 times | Joined on Jan 2010
#13
Great idea taixzo, I'll be sure to check it out too

Originally Posted by taixzo View Post
Unfortunately I do not have an N9, and I don't have the money to get one. If someone else is interested in getting it running though, it should be simoke in theory. Is pocketsphinx in the N9 repos?
I'm not 100% sure from your first post, but has Saera actually been released? If not, it could be eligible to be entered into one of the 5 main categories in the maemo.org coding competition which is due to start next week (Monday) onwards.. I think Saera could make a great candidate - apps like these usually do very well and we have a lot of N950s and N9s available as prizes.

However if it has been released we do have an "updates to existing applications" category, so if you do any major code changes or functional updates over the next 3 months I'd encourage you to enter Saera It doesn't cost anything to enter, and who knows maybe you'll win yourself a new device

So I'd definitely encourage you to check out the wiki and see if Saera qualifies
__________________
Follow me on my neglected twitter @kojacker

Cybot950 - Control a robot with your N9/N950
SMSPetFeeder - Build a Bluetooth/SMS dog feeder with Qt, N950, and arduino
Nerf950 - Use your N9/N950 to fire a Nerf gun

Last edited by kojacker; 2012-06-08 at 05:55.
 

The Following 8 Users Say Thank You to kojacker For This Useful Post:
Posts: 958 | Thanked: 3,426 times | Joined on Apr 2012
#14
Originally Posted by kojacker View Post
Great idea taixzo, I'll be sure to check it out too


I'm not 100% sure from your first post, but has Saera actually been released? If not, it could be eligible to be entered into one of the 5 main categories in the maemo.org coding competition which is due to start next week (Monday) onwards.. I think Saera could make a great candidate - apps like these usually do very well and we have a lot of N950s and N9s available as prizes.

However if it has been released we do have an "updates to existing applications" category, so if you do any major code changes or functional updates over the next 3 months I'd encourage you to enter Saera It doesn't cost anything to enter, and who knows maybe you'll win yourself a new device

So I'd definitely encourage you to check out the wiki and see if Saera qualifies
Thank you for that! I had not thought of submitting it, but as I haven't released it yet I will submit it to the competition.

FYI, your link points to the image rather than the page.
 

The Following 12 Users Say Thank You to taixzo For This Useful Post:
Posts: 291 | Thanked: 435 times | Joined on Apr 2011 @ RO
#15
I'm curious what is peoples pocketsphinx dictation aquracy {mine is very low)
and i couldn't find a simple solution to increase.
 
Posts: 958 | Thanked: 3,426 times | Joined on Apr 2012
#16
Originally Posted by SHARP66 View Post
I'm curious what is peoples pocketsphinx dictation aquracy {mine is very low)
and i couldn't find a simple solution to increase.
I'm working on this. Pocketsphinx's default model (hub4wsj) was generated from a large number of sentences from the Wall Street Journal. The problem here is that while this gives a good foundation of grammar, it has many words unlikely to be used in normal speech (such as 'Uzbekistan'). These other words make for poor accuracy. On the other hand, one can hand-build a list of words and a simple grammar that works well for voice commands, but not well for more general speech (as we want Saera to handle). I think the best option may be to make a new large dictionary, but create it from a more relevant data source, such as TMO for example. In the meantime I'm just using the default wsj model until I can replace it with something better.
 

The Following 5 Users Say Thank You to taixzo For This Useful Post:
Posts: 958 | Thanked: 3,426 times | Joined on Apr 2012
#17
Here is the source at the moment. Not much works with the AI; Saera currently recognizes the following things:
  • Hello
  • Test/Testing
  • Remind me X
  • Put on some music
  • What is the answer to life the universe and everything

I don't have hooks into the system yet, so neither reminders or music control actually work. Also, accuracy is fairly poor, for the reasons I just posted. Saera seems to run a good deal slower than the Pocketsphinx demo that the voice recognition is based on, so I will see what I can do to speed that up.

To run: extract zip on N900. In x-terminal, cd to that directory and run
Code:
python saera.py
Attached Files
File Type: zip saera.zip (55.1 KB, 595 views)
 

The Following 21 Users Say Thank You to taixzo For This Useful Post:
Posts: 47 | Thanked: 28 times | Joined on May 2010
#18
This is AMAZING to actually someone is doing this, i have a N9 but seriously how to compile or even edit that code, But I AM 100% eager to test this for the N9... If your interested im always available
 

The Following 2 Users Say Thank You to super-hero For This Useful Post:
Posts: 958 | Thanked: 3,426 times | Joined on Apr 2012
#19
Originally Posted by super-hero View Post
This is AMAZING to actually someone is doing this, i have a N9 but seriously how to compile or even edit that code, But I AM 100% eager to test this for the N9... If your interested im always available
It should not need compiling, as it is python code. Actually, it runs equally well on my desktop; if someone has an N9 and can install gstreamer0.10-pocketsphinx, pocketsphinx-utils and whatever the dictionary package is called (it's something like pocketsphinx-ln-en-hub4), this may run on Meego without modification.
 

The Following 8 Users Say Thank You to taixzo For This Useful Post:
Posts: 1,397 | Thanked: 2,126 times | Joined on Nov 2009 @ Dublin, Ireland
#20
You could try to combine it with Maemo Profiler, which is already capable of triggering a good number of actions.
 

The Following 5 Users Say Thank You to ivgalvez For This Useful Post:
Reply

Tags
saera, speech-to-text

Thread Tools

 
Forum Jump


All times are GMT. The time now is 09:06.