Hello Kempniu,
many thanks for developing xbmcvc, it works pretty well and it has great potential.
At the weekend I installed xbmcvc using a virtual machine running Ubuntu (this
seems to be the easiest way for all users), controlling XBMC local network instances
(PC and Raspberry Pi). Then i compiled it directly on my Linux PC and since today
it runs on my old Nokia N900, N9 will follow :-)
In the past I wrote a simple shell script myself which i called 'Poor man's Siri
resp. Google Now' (mee42.sh -
http://talk.maemo.org/showpost.php?s=f31...ostcount=1)
using a well known method for Google's Voice API as I encountered really bad results
using (pocket)sphinx regarding recognition accuracy. And now it is vice versa, switching
from recognition of sentences to recognition of a certain class of single words makes
a huge difference.
Instead of rewriting and extending my stuff in Python i will just test your C code :-)
Feedback:
1. Accuracy: This is something you can not solve on your side. I am not a native English
speaker, maybe that is the reason, 80-90% are not enough if you drop down the menus to home.
I tried removing everything i do not use from the model directory, i.e.
"CONTEXT TV T_V PROGRAMS SETTINGS MENU SHUFFLE UNSHUFFLE REPEAT OFF TWENTY
THIRTY FORTY FIFTY SIXTY SEVENTY EIGHTY SPELL"
2. Segmentation faults: Switching several times from XBMC to normal mode and back seems
to crash the application.
Feature request:
a. I would appreciate if you can add "FAVOURITES" to your code, I think the API supports this.
b. I understand xbmcvc as a client to the XBMC server. Maybe it is a good idea to implement an
XBMC addon which provides acoustic feedback using espeak etc telling the user where he is in
the XBMC tree (thinking of blind people).
Blues