I complied/installed sphinx 2.0.4 on several boxes but each time it fails to recognize speech. Upon running sphin2-simple, it loads sphinx2-continuous, there are pages of messages (verbosity=9), and it says "READY". Upon speaking into the microphone, it says:
Nothing happens after that point- as if it froze. I tried strace, and there are no strace messages after "READY". According to 'top', sphinx2-continuous uses almost 100% of CPU while loading, but practically no CPU when running. So, it appears it is doing nothing.
Sphinx2-test is sucessfull.
I tried this on an intel suse linux 8.0 box, and on an AMD suse linux 8.0 box. Both are linux 2.4.18-4gb. Both use ALSA. I tried it both with onboard ac97 audio, then I bought a SB Live 5.1 card, and verified the EMU10K1 driver was ok. None worked.
Microphone volume is up. I tried recording with arecord, and successfully made a recording comparable to the goforward.16k file. I havent seen any instructions on how to make sphinx read the file rather than the mic - if anyone can tell me how, this might be a good test.
I just checked /var/log/messages, and I did find something. I saw the follwing message appear when loading sphinx-continuous:
linux kernel: ALSA ../alsa-kernel/core/pcm_lib.c:186: Unexpected hw_pointer value (stream = 1, delta: -1168, max jitter = 2048): wrong interrupt acknowledge?
I though maybe Sphinx was trying to record at 44khz. Maybe ALSA is being weird. I'm pretty much stuck, and since I'm new at this, I'm not in much of a position to hack code. Anyone have ideas? Or does anyone have a SB Live/linux setup that works?
Thank you,
Jeremy
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I had a similar problem, which I tracked down to having words in the vocabulary that were not in the dictionary. When I reconciled the two (I have a Python script that does this for me), I was able to get recognition going.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I complied/installed sphinx 2.0.4 on several boxes but each time it fails to recognize speech. Upon running sphin2-simple, it loads sphinx2-continuous, there are pages of messages (verbosity=9), and it says "READY". Upon speaking into the microphone, it says:
READY....
INFO: uttproc.c(897): Livemode
INFO: agc_emax.c(72): AGCEMax: 5.00
Listening...
Nothing happens after that point- as if it froze. I tried strace, and there are no strace messages after "READY". According to 'top', sphinx2-continuous uses almost 100% of CPU while loading, but practically no CPU when running. So, it appears it is doing nothing.
Sphinx2-test is sucessfull.
I tried this on an intel suse linux 8.0 box, and on an AMD suse linux 8.0 box. Both are linux 2.4.18-4gb. Both use ALSA. I tried it both with onboard ac97 audio, then I bought a SB Live 5.1 card, and verified the EMU10K1 driver was ok. None worked.
Microphone volume is up. I tried recording with arecord, and successfully made a recording comparable to the goforward.16k file. I havent seen any instructions on how to make sphinx read the file rather than the mic - if anyone can tell me how, this might be a good test.
I just checked /var/log/messages, and I did find something. I saw the follwing message appear when loading sphinx-continuous:
linux kernel: ALSA ../alsa-kernel/core/pcm_lib.c:186: Unexpected hw_pointer value (stream = 1, delta: -1168, max jitter = 2048): wrong interrupt acknowledge?
I though maybe Sphinx was trying to record at 44khz. Maybe ALSA is being weird. I'm pretty much stuck, and since I'm new at this, I'm not in much of a position to hack code. Anyone have ideas? Or does anyone have a SB Live/linux setup that works?
Thank you,
Jeremy
I had a similar problem, which I tracked down to having words in the vocabulary that were not in the dictionary. When I reconciled the two (I have a Python script that does this for me), I was able to get recognition going.