Menu

problem in word recognition

Help
vijeta
2015-07-02
2015-07-02
  • vijeta

    vijeta - 2015-07-02

    Sir I am using precompiled binaries.
    When I run this command to recognize from file
    D:\pocketsphinx\bin\Debug>pocketsphinx_continuous -hmm model_parameters -lm punjabi.lm.DMP -dict punjabi.dic -infile Sound_2.wav
    then after giving the current configuration it gives the following error:
    INFO: feat.c(715): Initializing feature stream to type:

    '1s_c_d_dd', ceplen=13,CMN='current', VARNORM='no', AGC='none'
    INFO: cmn.c(143): mean[0]= 12.00, mean[1..12]= 0.0
    ERROR: "acmod.c", line 90: Folder 'model_parameters' does not contain acoustic model definition 'mdef'
    This is the link of model_parameters folder which I have used, which contains mdef.

    https://drive.google.com/file/d/0Bw_kmpUVdN3VdVpPSEp1d3lreUk/view?usp=sharing

     
    • Nickolay V. Shmyrev

      The error is in plain English. model_parameters folder does not have mdef file inside.

      The folder which has mdef is model_parameters\punjabi.cd_cont_200. So you need to run

          -hmm model_parameters\punjabi.cd_cont_200
      
       
  • vijeta

    vijeta - 2015-07-02

    Thank you so much sir, it works but gives now problem after this step as
    INFO: acmod.c(252): Parsed model-specific feature parameters from model_paramete
    rs\punjabi.cd_cont_200/feat.params
    INFO: feat.c(715): Initializing feature stream to type: '1s_c_d_dd', ceplen=13,
    CMN='current', VARNORM='no', AGC='none'
    INFO: cmn.c(143): mean[0]= 12.00, mean[1..12]= 0.0
    INFO: mdef.c(518): Reading model definition: model_parameters\punjabi.cd_cont_20
    0/mdef
    INFO: bin_mdef.c(181): Allocating 27107 * 8 bytes (211 KiB) for CD tree
    INFO: tmat.c(206): Reading HMM transition probability matrices: model_parameters
    \punjabi.cd_cont_200/transition_matrices

    And then a dialog box will appear which says pocketsphinx_continuous.exe has stopped working. Now is this a version problem of pocketsphinx because transition_matrices is present and might not be having any problem.
    I tried it a number of times yet it gives same problem.

     
    • bic-user

      bic-user - 2015-07-02

      You need to share acoustic model, language model, dictionary and audio file to reproduce your problem

       
  • vijeta

    vijeta - 2015-07-02

    Ok. This is the link of model_parameters folder.
    https://drive.google.com/file/d/0Bw_kmpUVdN3VdVpPSEp1d3lreUk/view?usp=sharing
    Remaining files are attached along this.

     
    • bic-user

      bic-user - 2015-07-02

      Works fine for me. Tried both precompiled prealpha and compiled from sources. Redo things from the very beginning and in case of failure share all possible information: log, exact error message, where you run executable from, etc.
      Btw, your file is to short, try recognizing something longer

       
  • vijeta

    vijeta - 2015-07-02

    Sir I am using pocketsphinx-5prealpha-win32.zip. Is that fine or I need to use pocketsphinx-5prealpha.tar.gz? I thought this can be the reason for problem.

     
    • bic-user

      bic-user - 2015-07-02

      win32 is for windows, tar.gz is for linux

       
  • vijeta

    vijeta - 2015-07-02

    Sir please tell me just the last thing. As if I wanted to show the recognized word to someone then how can I do that using this command? Because it is directly showing no such output.

     
    • bic-user

      bic-user - 2015-07-02

      Not sure I fully understand your question. Nothing is recognized in file you provide. It is too short and quite. For other files, to suppress log on windows you can add option:

      -logfn nul

       
  • vijeta

    vijeta - 2015-07-02

    No that's fine. That I will do for some other file. But now what I am asking is that if I consider it to be a college project and wanted to show the recognized word like to my teacher that I have spoken this word and it is recognized here. In that sense how can i do that.

     
  • bic-user

    bic-user - 2015-07-02

    I suppose you can show console with recognized word. If file recognition is not impressive enough you can run executable with

    -inmic yes

    instead of -infile some/file.wav. Input from microphone will be captured and recognized

     
  • vijeta

    vijeta - 2015-07-02

    ok thanks a lot sir for all your help.

     

Log in to post a comment.