Speech recognition with OCVolume

1080275535000

Downloaded a software called OCVolume - a voice recognition software. Before you can use you need to train it - which is the norm for most voice recognition software.

However the training program starts off on the wrong foot by giving the following pair of errors

SOUND_PCM_READ_CHANNELS: Invalid argument

SOUND_PCM_WRITE_CHANNELS: Invalid argument

As often happens with open source projects this one does not have adequate documentation. Unfortunately the project does not have a message board. Looks like i will have to figure this one out by myself.

This is not the only problem the OCVolume apparently seems to have. The binary version of the voice trainer only saves the last word that you work with - which i found to my great anoyance after working through about 25 words. Training is a tedius process because you have to save each word individually as a file.

I compiled the source code and ran it in debug mode only to find that it behaves completely differently from the source download. I think they may have accidentaly put up a wrong version of the source code.

I tried with both jre 1.3x and 1.4x with pretty much the same result. Great pity because the creators of OCVolume seem to have some exciting ideas.

comments powered by Disqus