Hello,
Followed all the steps from here: https://cmusphinx.github.io/wiki/tutorialpocketsphinx/
Built was successful in Visual Studio 2017.
Running pocketsphinx_continuous as shown for microphone input, does nothing. Program returns to DOS prompt right away. No messages, nothing. Both the Win32 and the 64 bit builds. Running Windows 10. Sound card input is present, microphone as well, as shown in Windows 10 settings.
Regards,
Vlad
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you! I thought the same, however there is already a dll by the required name in the exe folder by the end of the build process. Slightly different in size than the one that one is supposed to copy there. Anyway, turns out that if Windows is very loaded and busy, it just doesn't run programs. When several of the open windows and processes are closed, pocketsphinx_continuous starts working "normally".
One problem I notice is that it outputs a lot of information that scrolls on the command window, and it is hard to catch the result of the recognition. When I see a glimpse of the result, sometimes it is right, most of the time wrong. I was expecting better from a phoneme based recognizer. May be there is an option, or an #ifdef somewhere, if project has to be rebuilt, to avoid outputing so much text on the screen. Then probably it would be possible to at least visually assess the WER.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you! I thought the same, however there is already a dll by the
required name in the exe folder by the end of the build process. Slightly
different in size than the one that one is supposed to copy there. Anyway,
turns out that if Windows is very loaded and busy, it just doesn't run
programs. When several of the open windows and processes are closed,
pocketsphinx_continuous starts working "normally".
One problem I notice is that it outputs a lot of information that scrolls
on the command window, and it is hard to catch the result of the
recognition. When I see a glimpse of the result, sometimes it is right,
most of the time wrong. I was expecting better from a phoneme based
recognizer. May be there is an option, or an #ifdef somewhere, if project
has to be rebuilt, to avoid outputing so much text on the screen. Then
probably it would be possible to at least visually assess the WER.
Hello,
It works, but I would expect it to work better. In a sense it makes more errors than I thought it should make. And I couldn't find a way to remove all this text information scrolling, that may be important, but just clutters the output visually, and recognition result is burried inside. Ideally it should be possible to disable the display of anything else but the recognition result.
Best,
Vlad
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
Followed all the steps from here:
https://cmusphinx.github.io/wiki/tutorialpocketsphinx/
Built was successful in Visual Studio 2017.
Running pocketsphinx_continuous as shown for microphone input, does nothing. Program returns to DOS prompt right away. No messages, nothing. Both the Win32 and the 64 bit builds. Running Windows 10. Sound card input is present, microphone as well, as shown in Windows 10 settings.
Regards,
Vlad
It is probably missing the dlls which shoudl be in the same folder as exe file.
Thank you! I thought the same, however there is already a dll by the required name in the exe folder by the end of the build process. Slightly different in size than the one that one is supposed to copy there. Anyway, turns out that if Windows is very loaded and busy, it just doesn't run programs. When several of the open windows and processes are closed, pocketsphinx_continuous starts working "normally".
One problem I notice is that it outputs a lot of information that scrolls on the command window, and it is hard to catch the result of the recognition. When I see a glimpse of the result, sometimes it is right, most of the time wrong. I was expecting better from a phoneme based recognizer. May be there is an option, or an #ifdef somewhere, if project has to be rebuilt, to avoid outputing so much text on the screen. Then probably it would be possible to at least visually assess the WER.
Hi
Did you get it to work properly?
regards
On Sat, Sep 7, 2019 at 6:30 AM Vladimir Botchev vladobb@users.sourceforge.net wrote:
Hello,
It works, but I would expect it to work better. In a sense it makes more errors than I thought it should make. And I couldn't find a way to remove all this text information scrolling, that may be important, but just clutters the output visually, and recognition result is burried inside. Ideally it should be possible to disable the display of anything else but the recognition result.
Best,
Vlad
https://stackoverflow.com/questions/17825820/how-do-i-turn-off-e-info-in-pocketsphinx/