WavFrag

The Wavfrag voice recognition engine works on an entirely new principle. While the technical details can be found on the technology page, here we explain the controls that are associated with voice recognition.

Starting recognition is as easy as firing up the WaveFrag application. The application starts up armed, and it will listen as soon as it is initialized. Naturally, one can start / stop WaveFrag, with the buttons to the right. When WaveFrag is listening, and there is an active sound, the green LED lights up. As pictured on the right, the LED is idle, and it shows in dark gray.

The threshold slider. With the slider titled 'Tresh', one can adjust the trigger (volume) point of WaveFrag. When the slider is up, WaveFrag needs high a volume of sound to trigger it. When the slider is down, WaveFrag is triggered with the slightest noise. The slider needs to be adjusted, so that WaveFrag is triggered only when you speak. This slider is not sensitive, and apart from extreme cases the default position will yield good results. The slider setting will be remembered across sessions

The volume slider. With the slider titled 'Vol' one can adjust the microphone volume. WaveFrag is not very sensitive to the volume settings. Additionally, one can always use the windows built in mixer to adjust the volume of the microphone. In special cases one may need to adjust the microphone boost settings and the Windows mixer to get good results.

WaveFrag can accept feed from any of the sound input devices. If you want a different device to be WaveFrag's input, select the desired device in the 'Input:' drop-down box . WaveFrag starts up with the default device, and it will remember its previous setting. WaveFrag plays to the default output device, but one can set any output that's available on this computer.

WaveFrag has a special 'strict' mode. In this mode WaveFrag uses a stricter criteria to signal matches. This results in more accurate recognition, less false positives at the expense of

requiring a cleaner pronunciation. This mode may be appropriate for some uses, where false triggering is undesirable. in this mode WaveFrag signals a match only if the top two recognition dictionary look up match.