Menu

Segmentation Fault in Sphin3

Help
2003-04-22
2012-09-22
  • Carlos Menezes

    Carlos Menezes - 2003-04-22

    Does someone know what is hapening?

    When I try sphinx3-simple, I get a segmentation fault.....
    ............
    ............
    ............
    INFO: subvq.c(271): Reading subvq 0
    INFO: subvq.c(273): Reading codebook
    INFO: subvq.c(293): Reading map
    INFO: subvq.c(271): Reading subvq 1
    INFO: subvq.c(273): Reading codebook
    INFO: subvq.c(293): Reading map
    INFO: subvq.c(271): Reading subvq 2
    INFO: subvq.c(273): Reading codebook
    INFO: subvq.c(293): Reading map
    INFO: subvq.c(103): Precomputing Mahalanobis distance invariants
    INFO: tmat.c(135): Reading HMM transition probability matrices: /usr/local/share/sphinx3/model/hmm/hub4_cd_continuous_8gau_1s_c_d_dd/transition_matrices
    ERROR: "tmat.c", line 197: Normalization failed for tmat 2 from state 0
    ERROR: "tmat.c", line 197: Normalization failed for tmat 2 from state 1
    ERROR: "tmat.c", line 197: Normalization failed for tmat 2 from state 2
    INFO: tmat.c(217): Read 48 transition matrices of size 3x4
    INFO: wid.c(92): 6 LM words not in dictionary; ignored
    INFO: dict2pid.c(254): Building PID tables for dictionary
    INFO: dict2pid.c(422): 873 composite states; 324 composite sseq
    INFO: kbcore.c(172): Verifying models consistency:
    INFO: kb.c(141): Building lextrees
    INFO: kb.c(207): Lextrees(3), 1428 nodes(ug), 1 nodes(filler)
    INFO: kb.c(229): Beam= -460509, PBeam= -383758, WBeam= -268630, SVQBeam= -15350
    INFO: vithist.c(71): Initializing Viterbi-history module

    Hit <cr> BEFORE and AFTER speaking:
    INFO: feat.c(943): Feature buffers initialized to 256 vectors
    INFO: cmn_prior.c(72): mean[0]= 12.00, mean[1..12]= 0.0
    /usr/local/bin/sphinx3-simple: line 11: 21509 Segmentation fault      $S3CONTINUOUS ${ARGS}

    Thanks in advance,

    Carlos Menezes

     
    • Scott Silliman

      Scott Silliman - 2003-04-23

      yeah, I get the same exact error a on RH7.3 box and also on 2 RH8.0 machines.  What OS are you running?  I can get it to recognize utterances from pre-recorded files, but not from the microphone.

      Does anyone have microphone based recognition working on linux, or specifically RH? 

       
    • Carlos Menezes

      Carlos Menezes - 2003-04-23

      Scott,

      I am using Debian Linux (Woody, stable). I got recognition working with Sphinx 2.
      My problems are only with Sphinx 3.
      Any sugestion?

      Regards,

      Menezes

       
    • Scott Silliman

      Scott Silliman - 2003-04-23

      same here. sphinx2 works fine, but sphinx3 is only working on pre-recorded files.  Anyone out there experienced this and found a fix for it?

       
    • Carlos Menezes

      Carlos Menezes - 2003-04-23

      Scott,

      I sent the same message to forum "Open Discussion" and received the following answer:

      > By: jphekman ( Jessica P. Hekman )
      > RE: Segmentation Fault in Sphinx3 
      > 2003-04-23 16:12 I'm not sure if this is the actual problem, but you are getting the message that there are words in the LM which aren't in your dictionary. I'd try figuring out what they are and putting them in the dictionary, and see how that goes.
      >
      >Jessica

      I hope this sugestion will help us.

      Menezes

       
    • Scott Silliman

      Scott Silliman - 2003-04-23

      Menezes,

      I extracted the vocab list out of the text based lm (an4.ug.lm) and compared ti with an4.dict and found the following pronunciations missing:

      HALL               HH AO L
      LANE               L EY N
      MEMORY         M EH M ER IY
      TWELVTH        T W EH L V TH
      WEAN             W IY N

      adding these to the dictionary gets past the seg fault we were getting, but it doesn't get very far before it segfaults elsewhere

      Hit <cr> BEFORE and AFTER speaking:
      INFO: feat.c(943): Feature buffers initialized to 256 vectors
      INFO: cmn_prior.c(72): mean[0]= 12.00, mean[1..12]= 0.0
      INFO: live.c(239): live_nfeatvec: 21
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 25
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 26
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 25
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 26
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 26
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 25
      INFO: main_live_example.c(128): PARTIAL HYP: <sil>
      INFO: live.c(239): live_nfeatvec: 0
      ./sphinx3-simple: line 11: 15979 Segmentation fault      $S3CONTINUOUS ${ARGS}

      I might have to wait a few days before I have time to look into this further though...

      -Scott

       
    • datta prasad

      datta prasad - 2003-05-02

      dear all,
      i m new to sphinx3. i trained an acoustic model  using Sphinx Train and now i need to configure the acoustic model to Sphinx 3. i m getting trouble in configuring this. the args file asks for the path of means, variances , mixture weights and subvq. which path should i have to point to in that file.

      Pls help me in this regard..

      thanks in advance

       

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.