Active Topics
-
Error after installing Leste on sdcard (13)
to Maemo 7 / Leste by Maemish - 11 hrs, 40 mins ago -
TLS 1.2 for N9 is ready (29)
to MeeGo / Harmattan by smatkovi - 20 hrs, 29 mins ago -
Camera phone competition March 2023 "Endless" (35)
to General by Maemish - 4 days, 5 hrs ago -
Maemo repository (3)
to Community by torsen - 4 days, 22 hrs ago - more...
|
2009-12-11
, 10:49
|
|
Posts: 11 |
Thanked: 1 time |
Joined on Dec 2009
|
#12
|
definitely need voice dialling and speed dial. What i would also love to see is voice number recognition where I can speak numbers and the N900 would dial. Comes in handy when you have to take down a number while driving or when hands are busy. That's on my wishlist, but we definitely need voice dialing on a 'smartphone.'
|
|
2010-08-02
, 15:17
|
|
Posts: 3 |
Thanked: 0 times |
Joined on Jul 2010
@ UK
|
#13
|
It is quite extraordinary that Nokia permits the release of a flagship 'phone without voice activated dialling. In fact voice activated commands (such as for media selection) and voice commands for the Navigation are all essential if this device is to be of any use in a vehicle.
|
|
2010-08-02
, 15:26
|
|
Posts: 121 |
Thanked: 40 times |
Joined on Dec 2009
@ Tokyo, Japan
|
#14
|
Just today I've thought it would be so great to have voice dialling...maybe someone will develop it.
|
|
2010-08-02
, 15:32
|
|
Posts: 1,751 |
Thanked: 844 times |
Joined on Feb 2010
@ Sweden
|
#15
|
There are some projects on this in the linux world. I do not know how vivid they are but they do exist.
It would be cool but not necessary for me. I would prefer if it come from the community.. cause if it came from Nokia it would probably be a closed source blobb..
It would be cool but not necessary for me. I would prefer if it come from the community.. cause if it came from Nokia it would probably be a closed source blobb..
|
|
2010-10-13
, 08:09
|
|
Posts: 88 |
Thanked: 9 times |
Joined on Oct 2010
|
#16
|
Is anyone still thinking of developing a voice dialler as this would be very nice to have on the N900. Scrolling through contacts whilst it is in car holder is tricky and also illegal here in UK!
|
|
2010-10-13
, 10:11
|
|
Posts: 153 |
Thanked: 159 times |
Joined on Nov 2009
|
#17
|
I recently started researching the possibility to implement voice dialing.
What i (think) to know so far and whats my ideas / possible hickups are:
- Real speech recognition is hard to implement
- Assigning a wav file to a contact and then comparing that recording to the new recording should be easier to implement so thats the direction i used in my research.
I found a library and sample for fingerprinting a wave file to use for comparsion. (post the url and name later don't remember the name), but had not the time to test it if i create, lets say 5 wave recordings of me saying the same name and then fingerprinting the wav files and test if i get the same fingerprint.
So i thing first part of creating a working voice dialing application for maemo/meego should create or find a way to get some kind of speech recognition or speech comparing/fingerprinting.
After that my idea was to use n900 communication api to use. like adding a new communication service (like jabber, skype...) and then add a new field to each contact to assign a wave for the "voice dial" service account..
After the dialing itself i had two ideas.
First one is quite easy, you press the mic button, i catch the press and then play a short beep and then record from the mic the next 5 to 10 seconds and compare the recording / fingerprint with all stored contacts wav files, only drawbacks i guess is if you use some other apps which listens to the dbus event of pressing the mic button.
Other idea is to assign a "master word" wave recording to the communication service itself like "dial" and then if you say "dial" it recognize that you want to start the voice dialing and do the same as if you pressed the mic button in the first example.
The problems i see is that we always must record stuff and if its in a specific volume process it to detect we recorded the "master word". also a problem could be where we think the word began or ended. this also will reduce the battery life.
Thats only my ideas i had so far.
So my ideas in short and my preferred order and solutions:
1. Find or create a working wave fingerprinting/comparison algorithm
2. Implement an easy assign a wave to a contact method, either via a communication service plugin or a separate app.
3. Write an dbus listener which listens to to mic button press and then starts recording, comparing and if a match found play the match (for comparsion that you are sure you dial the right one) and then start dialing.
edit:
the software i found is called: libfooid and i use it from here:
http://blog.asymptotic.co.uk/softwar...ing-mp3-files/
currenlty doing some tests with it
Last edited by smurfy; 2010-10-13 at 18:37.
What i (think) to know so far and whats my ideas / possible hickups are:
- Real speech recognition is hard to implement
- Assigning a wav file to a contact and then comparing that recording to the new recording should be easier to implement so thats the direction i used in my research.
I found a library and sample for fingerprinting a wave file to use for comparsion. (post the url and name later don't remember the name), but had not the time to test it if i create, lets say 5 wave recordings of me saying the same name and then fingerprinting the wav files and test if i get the same fingerprint.
So i thing first part of creating a working voice dialing application for maemo/meego should create or find a way to get some kind of speech recognition or speech comparing/fingerprinting.
After that my idea was to use n900 communication api to use. like adding a new communication service (like jabber, skype...) and then add a new field to each contact to assign a wave for the "voice dial" service account..
After the dialing itself i had two ideas.
First one is quite easy, you press the mic button, i catch the press and then play a short beep and then record from the mic the next 5 to 10 seconds and compare the recording / fingerprint with all stored contacts wav files, only drawbacks i guess is if you use some other apps which listens to the dbus event of pressing the mic button.
Other idea is to assign a "master word" wave recording to the communication service itself like "dial" and then if you say "dial" it recognize that you want to start the voice dialing and do the same as if you pressed the mic button in the first example.
The problems i see is that we always must record stuff and if its in a specific volume process it to detect we recorded the "master word". also a problem could be where we think the word began or ended. this also will reduce the battery life.
Thats only my ideas i had so far.
So my ideas in short and my preferred order and solutions:
1. Find or create a working wave fingerprinting/comparison algorithm
2. Implement an easy assign a wave to a contact method, either via a communication service plugin or a separate app.
3. Write an dbus listener which listens to to mic button press and then starts recording, comparing and if a match found play the match (for comparsion that you are sure you dial the right one) and then start dialing.
edit:
the software i found is called: libfooid and i use it from here:
http://blog.asymptotic.co.uk/softwar...ing-mp3-files/
currenlty doing some tests with it
Last edited by smurfy; 2010-10-13 at 18:37.
|
|
2010-10-13
, 10:46
|
|
Posts: 1,397 |
Thanked: 2,126 times |
Joined on Nov 2009
@ Dublin, Ireland
|
#18
|
A nice project that could be ported to Maemo to accomplish this task is Simon Listens for KDE.
I have also seen that Sphinx is now in the repos with a lot of libraries, so someone is maybe working on this.
I have also seen that Sphinx is now in the repos with a lot of libraries, so someone is maybe working on this.
Voice dialing is just the tip of the iceberg.