CODEWITHSUNDEEP
osx-snow-leopardspeech-recognitionhtk
oatawa1
oatawa1

Reputation: 21

Live recognition via HTK on OS X ERROR [+6306] OpenAsChannel: Audio input not supported

My objective is doing live recognition via HTK 3.4.1 on OS X.

First, I did it on Ubuntu 10.0.4. After I generated acoustic and language model, also everything it need. I execute this command to do live recognition.

    HVite -H am/tiehmm2m_5/newMacros \
          -C config/liverecog.config \
          -w lm/dgs.wdnet \
           config/dgs.dict \
           config/tie.list

Everything work fine for me. Then I try to port it to run on my OS X, with the same code. I can do everything such as generate model or recognition.

But I got an error like this when I did live recognition.

    READY[1]>
     ERROR [+6306]  OpenAsChannel: Audio input not supported
     FATAL ERROR - Terminating program HVite

How can I solve this error? Anyone face this error before?

Thanks in advance.

Upvotes: 1

Views: 289

Answers (1)

Nikolay Shmyrev
Nikolay Shmyrev

Reputation: 25220

It says that input on OSX is not supported. Not supported means that some feature is not implemented in the code. If you need this feature you have to implement it or look for another tool which fits your needs.

The reason that OSX is different is that audio input subsystem in OSX (CoreAudio) is quite different from the one from Linux (OSS). You just need to implement audio input on OSX from scratch.

Upvotes: 1

Related Questions