VoxForge
I created an Australian English Acoustic Model based on the ANDOSL corpus with the already compiled HTK 3.3 components. The recognition with HVite.exe works brilliant.
Now, as I wanted to integrate HTK into an application network I downloaded the sourcecode from HTK website (only 3.4.1 available), made my changes and compiled it.
But when I run HVite 3.4.1 I get the error:
ERROR [+6306] OpenAsChannel: Audio input not supported
Same with the precompiled 3.4 version from the HTK website.
Someone had the same problem, and solved it or got some good advice?
Kind Regards,
UC
Using HDecode is not an option at the moment because it doesn't support my acoustic model and at the moment I don't want to create a new one using HDecode.
--- (Edited on 1/14/2013 10:23 pm [GMT-0600] by Visitor) ---
Hey,
I figured out that it doesn't have anything to do with the version 3.4.
I compiled the version 3.3 and the same problem occures. So something goes wrong while compile the code. I use the unchanged HTK HVite code with Visual Studio 2008. I create a new project like in the tutorial on (http://wangpidong.blogspot.com.au/2010/02/how-to-compile-htk-under-visual-studio.html).
Does someone has an idea why this error occures?
--- (Edited on 2/3/2013 6:35 pm [GMT-0600] by Visitor) ---
It just says what it says. It says HTK was compiled without WIN32 audio input and audio input is not supported.
You have several options:
1. Use input from a file instead of input from a microphone
2. Compile HTK with Win32 audio input. For that you need to define WIN32_AUDIO in project preprocessor flags in project properties. Probably you didn't define it in your project.
--- (Edited on 2/4/2013 06:22 [GMT+0300] by nsh) ---
Thank you very much. That solved my problem.
I added:
/wd4335 /D "WIN32_AUDIO" /D ARCH=\"WIN32\" /D isnan=_isnan
to Properties -> C/C++ -> Command Line and
winmm.lib to Properties -> Linker -> Input -> Additional Dependencies
Now it works properly with a microphone!
--- (Edited on 2/4/2013 12:03 am [GMT-0600] by Visitor) ---