I've created a Voice Recognition app, my aim is to create a simple Voice Recognition Interface that is easy change and reduces false recognised commands. And to allow you to configure the setup the way you want it.
So I've taken the features I like from NaviVoice & Bruno and built my own Voice recognition app:
SpeakEasy VR.
SpeakEasy VR unique features:
1. Highly customisable configuration
2. Predictive help to complete command
3. Interactively set a acceptance threshold for valid recognised commands
4. Simple interface to change voice commands (car friendly resolution)
5. Locking of command groups also enable/disable comand groups by voice
6. Can control and application that has a SendMessage, HotKey or TCPIP interface. I'll make a Girder importer eventually if it is popular.
7. Logging of voice commands to figure out which commands work well or badly in you car (with a noisey car engine a Voice Recogniser can pick up false commands). So this is a good way to figure out if you need to alter some voice phrases.
Ignore my orange/black RR skin that's my own funky colored RR skin. The window to look at is the slightly transparent white bix on the top left and in the center. The top left box is clickable to start the engine. The center box displays while you are talking. You can voice activate the center box as well and set the transparency levels to your liking.
I have a worked Beta if anyone is brave enough to try it!
----------------------------------
BTW, Interpreted text in GREEN means that it was recognised the accuracy was ABOVE the threshold which you set in the slider. If the text is RED the it was recognised but BELOW the threshold.
The Threshold is between 0% and 50%. Setting it to 0% means that anything that is recognised is executed. Set the value higher mean that you have to speak more clearly and precisely. I have a Irish accent and it seems to not work so well with it so I set the threshold high to about 25%, it also reduces false recognised phrases produced by car noise and MP3 playback. As you speak you'll see the accuracy of each phrase that is recognised, try to figure out what accuracy works best for you and set the slider threshold level to just below what you normally get for the accuracy. I find that for some phrases I need to turn it way down so that it is recognised, while other phrases get close to 100% no problem.
I'll write up some documentation in the next release if there aren't too many bugs/crashes
