Menu

demos extremely slow

Help
bill
2005-03-21
2012-09-22
  • bill

    bill - 2005-03-21

    I am attempting to familiarize myself with sphinx in hopes of adding speech recognition to an Electronic Medical Records program I have written.

    I am running Linux Mandrake 10.0. The demos are extremely slow (for example - 5? minutes before the opening screen on HelloWorld comes up). strace indicates that repeated system calls are going out over and over again for /dev/snd/controlCxx. It is making other repetitive system calls as well.

    Any suggestions? Thanx. Bill

     
    • Willie Walker

      Willie Walker - 2005-03-21

      Hi Bill:

      Let's see if we can get you going! Can you please tell us more about your system:

      o What version of the Java platform you are using
      o What kind of hardware (RAM, CPU, etc) you have
      o How you are running the HelloWorld application

      Thanks!

      Will

      PS - If you are not already doing so, you might try running HelloWorld with the "-mx" option as described in the README.html file in the demo/sphinx/helloworld directory:

      java -mx312m -jar bin/HelloWorld.jar

       
    • bill

      bill - 2005-03-22

      Will:

      I am running java 1.42SE
      2.4 gigahertz processor 256MB ram

      I am running from the command line
      java -mx256m -jar bin/HelloWorld.jar (I think my readme says 256 not 312); however, I have the same problem with
      the other demos that do not suggest setting ?memory allocation? . It is not a java problem per se, i can run the non-sphinx demos from java. I also have no difficulty running the speakers and mircrophone through other programs.

      Thank you for your time.

      Bill

       
      • Willie Walker

        Willie Walker - 2005-03-22

        Hi Bill:

        When I run the HelloWorld demo as follows:

        java -mx256m -jar bin/HelloWorld.jar

        It takes 30 seconds for J2SE 1.4.2_05 on my 867MHz PowerBook to get from the following:

        java -mx256m -jar bin/HelloWorld.jar
        Loading...

        To the following (which is where the HelloWorld application is waiting for me to say something):

        Say: (Good morning | Hello) ( Bhiksha | Evandro | Paul | Philip | Rita | Will )
        Start speaking. Press Ctrl-C to quit.

        HelloWorld is a "live" application, waiting for you to say something into the microphone.

        When I re-read your first post, it looks as though Sphinx-4 really is listening for input, but maybe it's not hearing you. That is, perhaps your microphone is not set up such that Sphinx-4 can get to it. You might try running the AudioTool demo to see if your speech is actually making it to Sphinx-4.

        If you're still seeing very slow times, I'm going to guess that you may actually need more RAM (256M is a tight space to share).

        Will

         
    • bill

      bill - 2005-03-22

      Will

      When I run HelloWorld it takes a maximum of 43.5% of the ram. It is often taking less than 1% of the ram. The program takes 2min 31sec to go from loading... to say:(Good Morning... I have never actually gotten any evidence that microphone input is being received by the program. That could be attributed to java making no connection to the microphone or to the program running so slowly that nothing ever appears on the screen.

      When I run AudioTool, I again get no evidence that the microphone is actually connecting to the program. I do get the opening screen. When I click on RECORD, etc. I do get a change in the button that indicates it has been depressed but again the response is very slow.

      Thanks again for your time.

      Output of java -jar bin/AudioTool.jar -dumpMixers follows:

      Mixer[0]: "Java Sound Audio Engine"
      Description: Software mixer and synthesizer
      SourceLineInfo (e.g., speakers):
      PCM_SIGNED, -1.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, -1.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, -1.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, -1.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, -1.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, -1.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, -1.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, -1.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      TargetLineInfo (e.g., microphones):
      none
      Mixer[1]: "I82801DBICH4 [plughw:0,0]"
      Description: Intel 82801DB-ICH4, Intel ICH, Intel 82801DB-ICH4
      SourceLineInfo (e.g., speakers):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      TargetLineInfo (e.g., microphones):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      Mixer[2]: "I82801DBICH4 [plughw:0,1]"
      Description: Intel 82801DB-ICH4, Intel ICH - MIC ADC, Intel 82801DB-ICH4 - MIC ADC
      SourceLineInfo (e.g., speakers):
      none
      TargetLineInfo (e.g., microphones):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      Mixer[3]: "I82801DBICH4 [plughw:0,2]"
      Description: Intel 82801DB-ICH4, Intel ICH - MIC2 ADC, Intel 82801DB-ICH4 - MIC2 ADC
      SourceLineInfo (e.g., speakers):
      none
      TargetLineInfo (e.g., microphones):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      Mixer[4]: "I82801DBICH4 [plughw:0,3]"
      Description: Intel 82801DB-ICH4, Intel ICH - ADC2, Intel 82801DB-ICH4 - ADC2
      SourceLineInfo (e.g., speakers):
      none
      TargetLineInfo (e.g., microphones):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      Mixer[5]: "I82801DBICH4 [plughw:0,4]"
      Description: Intel 82801DB-ICH4, Intel ICH - IEC958, Intel 82801DB-ICH4 - IEC958
      SourceLineInfo (e.g., speakers):
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, -1.0 Hz, 16 bit, stereo, little-endian, audio data
      TargetLineInfo (e.g., microphones):
      none
      Mixer[6]: "Linux,dev/dsp,multi threaded"
      Description: No details available
      SourceLineInfo (e.g., speakers):
      none
      TargetLineInfo (e.g., microphones):
      ULAW, 8000.0 Hz, 8 bit, mono, audio data
      ULAW, 11025.0 Hz, 8 bit, mono, audio data
      ULAW, 16000.0 Hz, 8 bit, mono, audio data
      ULAW, 22050.0 Hz, 8 bit, mono, audio data
      ULAW, 32000.0 Hz, 8 bit, mono, audio data
      ULAW, 44100.0 Hz, 8 bit, mono, audio data
      ULAW, 8000.0 Hz, 8 bit, stereo, audio data
      ULAW, 11025.0 Hz, 8 bit, stereo, audio data
      ULAW, 16000.0 Hz, 8 bit, stereo, audio data
      ULAW, 22050.0 Hz, 8 bit, stereo, audio data
      ULAW, 32000.0 Hz, 8 bit, stereo, audio data
      ULAW, 44100.0 Hz, 8 bit, stereo, audio data
      ALAW, 8000.0 Hz, 8 bit, mono, audio data
      ALAW, 11025.0 Hz, 8 bit, mono, audio data
      ALAW, 16000.0 Hz, 8 bit, mono, audio data
      ALAW, 22050.0 Hz, 8 bit, mono, audio data
      ALAW, 32000.0 Hz, 8 bit, mono, audio data
      ALAW, 44100.0 Hz, 8 bit, mono, audio data
      ALAW, 8000.0 Hz, 8 bit, stereo, audio data
      ALAW, 11025.0 Hz, 8 bit, stereo, audio data
      ALAW, 16000.0 Hz, 8 bit, stereo, audio data
      ALAW, 22050.0 Hz, 8 bit, stereo, audio data
      ALAW, 32000.0 Hz, 8 bit, stereo, audio data
      ALAW, 44100.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 8000.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 8000.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 11025.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 11025.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 16000.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 16000.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 22050.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 22050.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 32000.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 32000.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 44100.0 Hz, 8 bit, mono, audio data
      PCM_UNSIGNED, 44100.0 Hz, 8 bit, mono, audio data
      PCM_SIGNED, 8000.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 8000.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 11025.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 11025.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 16000.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 16000.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 22050.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 22050.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 32000.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 32000.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 44100.0 Hz, 8 bit, stereo, audio data
      PCM_UNSIGNED, 44100.0 Hz, 8 bit, stereo, audio data
      PCM_SIGNED, 8000.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 8000.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 11025.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 11025.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 16000.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 16000.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 22050.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 22050.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 32000.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 32000.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 44100.0 Hz, 16 bit, mono, big-endian, audio data
      PCM_SIGNED, 44100.0 Hz, 16 bit, mono, little-endian, audio data
      PCM_SIGNED, 8000.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 8000.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, 11025.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 11025.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, 16000.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 16000.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, 22050.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 22050.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, 32000.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 32000.0 Hz, 16 bit, stereo, little-endian, audio data
      PCM_SIGNED, 44100.0 Hz, 16 bit, stereo, big-endian, audio data
      PCM_SIGNED, 44100.0 Hz, 16 bit, stereo, little-endian, audio data

       
      • Willie Walker

        Willie Walker - 2005-03-22

        Hi Bill:

        When you run AudioTool, it should be quite quick. The expected behavior with AudioTool is as follows:

        1) Press "Record" button, it's text should change to "Stop."
        2) Speak into the microphone.
        3) Press the "Stop" button. After a small delay, the button's text should change to "Record", you should hear the audio you just spoke, and you should also see the spectrogram.

        If you're not getting this, then something is whacked with the audio. You might try selecting a different mixer as describe in http://cmusphinx.sourceforge.net/sphinx4/doc/Sphinx4-faq.html#microphone_selection.

        Will

         
    • bill

      bill - 2005-03-22

      Will,

      I have been playing with AudioTool. The program works but again very slowly. For example, I press play, count from 1 to 85, the stop button comes up, I press stop, wait a few seconds and then 84, 85 comes out of the speaker.

      When I set -Dmicrophone[selectMixer]=1 through 5, I get a line unsupported error and the program does not respond even as poorly as above.

      Thanx Bill

       
      • Willie Walker

        Willie Walker - 2005-03-22

        Darn. Well....my last resort is to see if you are using Sun's JDK? If you are using that, then I'm at a loss for what could be going wrong here.

        Will

         
    • bill

      bill - 2005-03-23

      I agree, Darn.

      I am using the Sun JDK.

      I am going to install SUSE 9.0 linux and start over again.
      Mandrake is not listed as one of the supported linuxes by sun.

      I'll keep everyone posted. Thanks again. Bill

       
    • bill

      bill - 2005-03-24

      I did install SUSE 9.1 and now AudioTool works.

      I'm not sure if this was a problem with Mandrake or with my configuration of Mandrake but SUSE may be the way to go.

      Thanks for all your help.

      Bill

       

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.