» Tagged as: voice
Finally managed to get the Sphinx 4 demos to run with out throwing a heap of exceptions. This success sadly was not achieved on my linux box.
After deleting the entire source tree followed by another CVS checkout, the build process went smoothly and the programs actually started up without any error. Unfortunately there is still a gap between the 'saying and the doing'
Though some of the demo programs consumed upto 468 mega bytes of memory they simply refused to respond to voice input. The README that accompanies the demo suggests using the AudioTool to ensure that java programs are capable of producing or capturing sound on your linux system. The fate that befell AudioTool is suffered the same fate but it was capable of playing back sound from a disk file.
In any case I doubt if my audio set up is the culprit here (even though ALSA on linux does have quite a few tricks up it's sleeve). Afraid of completely going bold, I reluctantly decided to try this out on a windows box. Guess what? all the demos started up and actually began to respond to voice input.
However there is still another step between responding to audio input and recognizing them but that's to be expected. A few more years will pass before we have Speech Recognition software capable of recognizing input from any user (random users) without any prior training.