• Post Reply
  • Bookmark Topic Watch Topic
  • New Topic

Voice recognition in J2ME

 
Ransika deSilva
Ranch Hand
Posts: 524
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi,
I have this idea to do a project like this. It is a Global Phone book. The server has a list of phone numbers with the names or service. The name or the service name is saved as a bit pattern which is generated according to voice. For instance there can be a list of numbers for a taxi, when the phone user says 'taxi' to the MIDlet running on the phone, the MIDlet captures the voice, converts it in to some kind of a bit pattern, then sends it to the web application. The web application then matches the bit pattern and sends back the result.

I just wanted to know about your expertiese views on this, Is this viable with J2ME, MIDP 2.0?

Thanks a lot in advance....
 
Eduardo Marques
Ranch Hand
Posts: 231
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi,

MMAPI 1.1 is a J2ME optional package that allows sound recording.
(java.sun.com/products/mmapi)

As far as speech recognition is concerned, the Java Speech API
is not for J2ME, but there are several different implementations you can look at, including the FreeTTS project at Sourceforge.
Check links to FreeTTS and other implementations here:

http://java.sun.com/products/java-media/speech/
 
Ransika deSilva
Ranch Hand
Posts: 524
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi,
Thanks a lot for the reply, I will check the links, well I certainly would like to hear from you guys about the idea I am having, do you think it is possible, Thanks a lot...
 
William Frantz
Ranch Hand
Posts: 102
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
A more general system might be to avoid using a MIDlet altogether. Have the user call a voice mail type system attached to your server. Then speak the name ("Taxi"). The server could do the voice encoding, recognition and either read the results back over the phone (using text to speech) or send results back as a text message based on the caller ID of the user.

Such a system would operate much faster than encoding on the handset and uploading to a server. You also gain the benefit of immediate feedback for unrecognized words. Imagine the frustration of a user saying "Taxi" into a phone, waiting several minutes for it to upload to your server, then getting back some kind of "unrecognized, try again" type of response.

Besides, many (most?) phones will not include JSRs that allow access to the voice recorder.
 
Punit Raizada
Ranch Hand
Posts: 156
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
The concept is kinda cool... its going to be tough to implement however.. I dont have any clue about voice recognition so think its pretty tough and dependent on the surrounding noise etc.

But, have u thought about throwing in GPS position locator in the mix along with your application. If the midlet would know the position of the user and then provide the number of taxi service nearest to the user. I dont know any API's in J2ME that allow you to get the location of the user. Have seen some phones in BREW that allow you to do so.

Is there any API any J2ME that would allow the determining of the GPS location ?
 
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic