Menu ▾ ▴

Wierd pocketsphinx accuracy degredation.

Help
Jon
2012-08-07
2012-09-22

  • Jon

    Jon - 2012-08-07

    Pocketsphinx is producing a bizare issue where the accuracy appears to be
    degrading after only a couple of queries to the engine. The first query has
    near perfect accuracy - it can recognize relatively complicated and convoluted
    phrases without dificulty. However the second, and third recognitions can
    barely pick up a two sylable word, and by the forth query to the engine, it
    simply fails to generate a hypothesis.

    For context I'm creating a pocketsphinx application for android, using jsgf
    grammars (though the problem persists with fsg grammars as well). My code is
    based on the pocketsphinx demo for android
    http://cmusphinx.sourceforge.net/2011/05/building-pocketsphinx-on-
    android/    .

    I'm not quite sure whats causing the problem, the original demo appeared to
    work fine, and I have hardly altered the configuration of the speech engine,
    other than using jsgf grammars.

    Anyways, any one else ever experience something like this, or even have a
    sugestion as to what I could try to remedy this?

    Thanks

     

  • Nickolay V. Shmyrev

    Nickolay V. Shmyrev - 2012-08-08

    Provide pocketsphinx log which is created on the device

    Add

    -rawlogdir <somefolder>

    option to pocketsphinx initialization and collect the audio you are trying to
    recognize. Share the audio, maybe it's corrupted somehow.

     

  • Jon

    Jon - 2012-08-08

    Heres the raw data files. http://speechweb2.cs.uwindsor.ca/rawdata.zip

    Also heres my log 

    INFO: cmd_ln.c(691): Parsing command line:


Current configuration:
[NAME]      [DEFLT]     [VALUE]
-agc        none        none
-agcthresh  2.0     2.000000e+00
-alpha      0.97        9.700000e-01
-ascale     20.0        2.000000e+01
-aw     1       1
-backtrace  no      no
-beam       1e-48       1.000000e-48
-bestpath   yes     yes
-bestpathlw 9.5     9.500000e+00
-bghist     no      no
-ceplen     13      13
-cmn        current     current
-cmninit    8.0     8.0
-compallsen no      no
-debug              0
-dict               
-dictcase   no      no
-dither     no      no
-doublebw   no      no
-ds     1       1
-fdict              
-feat       1s_c_d_dd   1s_c_d_dd
-featparams         
-fillprob   1e-8        1.000000e-08
-frate      100     100
-fsg                
-fsgusealtpron  yes     yes
-fsgusefiller   yes     yes
-fwdflat    yes     yes
-fwdflatbeam    1e-64       1.000000e-64
-fwdflatefwid   4       4
-fwdflatlw  8.5     8.500000e+00
-fwdflatsfwin   25      25
-fwdflatwbeam   7e-29       7.000000e-29
-fwdtree    yes     yes
-hmm                
-input_endian   little      little
-jsgf               
-kdmaxbbi   -1      -1
-kdmaxdepth 0       0
-kdtree             
-latsize    5000        5000
-lda                
-ldadim     0       0
-lextreedump    0       0
-lifter     0       0
-lm             
-lmctl              
-lmname     default     default
-logbase    1.0001      1.000100e+00
-logfn              
-logspec    no      no
-lowerf     133.33334   1.333333e+02
-lpbeam     1e-40       1.000000e-40
-lponlybeam 7e-29       7.000000e-29
-lw     6.5     6.500000e+00
-maxhmmpf   -1      -1
-maxnewoov  20      20
-maxwpf     -1      -1
-mdef               
-mean               
-mfclogdir          
-min_endfr  0       0
-mixw               
-mixwfloor  0.0000001   1.000000e-07
-mllr               
-mmap       yes     yes
-ncep       13      13
-nfft       512     512
-nfilt      40      40
-nwpen      1.0     1.000000e+00
-pbeam      1e-48       1.000000e-48
-pip        1.0     1.000000e+00
-pl_beam    1e-10       1.000000e-10
-pl_pbeam   1e-5        1.000000e-05
-pl_window  0       0
-rawlogdir          
-remove_dc  no      no
-round_filters  yes     yes
-samprate   16000       1.600000e+04
-seed       -1      -1
-sendump            
-senlogdir          
-senmgau            
-silprob    0.005       5.000000e-03
-smoothspec no      no
-svspec             
-tmat               
-tmatfloor  0.0001      1.000000e-04
-topn       4       4
-topn_beam  0       0
-toprule            
-transform  legacy      legacy
-unit_area  yes     yes
-upperf     6855.4976   6.855498e+03
-usewdphones    no      no
-uw     1.0     1.000000e+00
-var                
-varfloor   0.0001      1.000000e-04
-varnorm    no      no
-verbose    no      no
-warp_params            
-warp_type  inverse_linear  inverse_linear
-wbeam      7e-29       7.000000e-29
-wip        0.65        6.500000e-01
-wlen       0.025625    2.562500e-02

INFO: cmd_ln.c(691): Parsing command line:
\
    -nfilt 20 \
    -lowerf 1 \
    -upperf 4000 \
    -wlen 0.025 \
    -transform dct \
    -round_filters no \
    -remove_dc yes \
    -svspec 0-12/13-25/26-38 \
    -feat 1s_c_d_dd \
    -agc none \
    -cmn current \
    -cmninit 56,-3,1 \
    -varnorm no

Current configuration:
[NAME]      [DEFLT]     [VALUE]
-agc        none        none
-agcthresh  2.0     2.000000e+00
-alpha      0.97        9.700000e-01
-ceplen     13      13
-cmn        current     current
-cmninit    8.0     56,-3,1
-dither     no      no
-doublebw   no      no
-feat       1s_c_d_dd   1s_c_d_dd
-frate      100     100
-input_endian   little      little
-lda                
-ldadim     0       0
-lifter     0       0
-logspec    no      no
-lowerf     133.33334   1.000000e+00
-ncep       13      13
-nfft       512     512
-nfilt      40      20
-remove_dc  no      yes
-round_filters  yes     no
-samprate   16000       8.000000e+03
-seed       -1      -1
-smoothspec no      no
-svspec             0-12/13-25/26-38
-transform  legacy      dct
-unit_area  yes     yes
-upperf     6855.4976   4.000000e+03
-varnorm    no      no
-verbose    no      no
-warp_params            
-warp_type  inverse_linear  inverse_linear
-wlen       0.025625    2.500000e-02

INFO: acmod.c(242): Parsed model-specific feature parameters from /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/feat.params
INFO: feat.c(684): Initializing feature stream to type: '1s_c_d_dd', ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean[0]= 12.00, mean[1..12]= 0.0
INFO: acmod.c(163): Using subvector specification 0-12/13-25/26-38
INFO: mdef.c(520): Reading model definition: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/mdef
INFO: mdef.c(531): Found byte-order mark BMDF, assuming this is a binary mdef file
INFO: bin_mdef.c(330): Reading binary model definition: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/mdef
INFO: bin_mdef.c(507): 50 CI-phone, 143047 CD-phone, 3 emitstate/phone, 150 CI-sen, 5150 Sen, 27135 Sen-Seq
INFO: tmat.c(205): Reading HMM transition probability matrices: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/transition_matrices
INFO: acmod.c(117): Attempting to use SCHMM computation module
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/means
INFO: ms_gauden.c(292): 1 codebook, 3 feature, size: 
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/variances
INFO: ms_gauden.c(292): 1 codebook, 3 feature, size: 
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(354): 0 variance values floored
INFO: s2_semi_mgau.c(908): Loading senones from dump file /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/sendump
INFO: s2_semi_mgau.c(932): BEGIN FILE FORMAT DESCRIPTION
INFO: s2_semi_mgau.c(1027): Using memory-mapped I/O for senones
INFO: s2_semi_mgau.c(1304): Maximum top-N: 4 Top-N beams: 0 0 0
INFO: dict.c(306): Allocating 137127 * 20 bytes (2678 KiB) for word entries
INFO: dict.c(321): Reading main dictionary: /mnt/sdcard/speechbrowser/temp.dict
INFO: dict.c(212): Allocated 1004 KiB for strings, 1657 KiB for phones
INFO: dict.c(324): 133021 words read
INFO: dict.c(330): Reading filler dictionary: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/noisedict
INFO: dict.c(212): Allocated 0 KiB for strings, 0 KiB for phones
INFO: dict.c(333): 11 words read
INFO: dict2pid.c(396): Building PID tables for dictionary
INFO: dict2pid.c(404): Allocating 50^3 * 2 bytes (244 KiB) for word-initial triphones
INFO: dict2pid.c(131): Allocated 30200 bytes (29 KiB) for word-final triphones
INFO: dict2pid.c(195): Allocated 30200 bytes (29 KiB) for single-phone word triphones
INFO: fsg_search.c(145): FSG(beam: -1080, pbeam: -1080, wbeam: -634; wip: -26, pip: 0)
INFO: jsgf.c(546): Defined rule: PUBLIC <test.hi>
INFO: fsg_model.c(213): Computing transitive closure for null transitions
INFO: fsg_model.c(264): 0 null transitions added
INFO: fsg_model.c(411): Adding silence transitions for <sil> to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++NOISE++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++BREATH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++SMACK++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++COUGH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++LAUGH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++TONE++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++UH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++UM++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_search.c(364): Added 13 alternate word transitions
INFO: fsg_lextree.c(108): Allocated 5406 bytes (5 KiB) for left and right context phones
INFO: fsg_lextree.c(251): 715 HMM nodes in lextree (586 leaves)
INFO: fsg_lextree.c(253): Allocated 77220 bytes (75 KiB) for all lextree nodes
INFO: fsg_lextree.c(256): Allocated 63288 bytes (61 KiB) for lextree leafnodes
INFO: cmd_ln.c(691): Parsing command line:


Current configuration:
[NAME]      [DEFLT]     [VALUE]
-agc        none        none
-agcthresh  2.0     2.000000e+00
-alpha      0.97        9.700000e-01
-ascale     20.0        2.000000e+01
-aw     1       1
-backtrace  no      no
-beam       1e-48       1.000000e-48
-bestpath   yes     yes
-bestpathlw 9.5     9.500000e+00
-bghist     no      no
-ceplen     13      13
-cmn        current     current
-cmninit    8.0     8.0
-compallsen no      no
-debug              0
-dict               
-dictcase   no      no
-dither     no      no
-doublebw   no      no
-ds     1       1
-fdict              
-feat       1s_c_d_dd   1s_c_d_dd
-featparams         
-fillprob   1e-8        1.000000e-08
-frate      100     100
-fsg                
-fsgusealtpron  yes     yes
-fsgusefiller   yes     yes
-fwdflat    yes     yes
-fwdflatbeam    1e-64       1.000000e-64
-fwdflatefwid   4       4
-fwdflatlw  8.5     8.500000e+00
-fwdflatsfwin   25      25
-fwdflatwbeam   7e-29       7.000000e-29
-fwdtree    yes     yes
-hmm                
-input_endian   little      little
-jsgf               
-kdmaxbbi   -1      -1
-kdmaxdepth 0       0
-kdtree             
-latsize    5000        5000
-lda                
-ldadim     0       0
-lextreedump    0       0
-lifter     0       0
-lm             
-lmctl              
-lmname     default     default
-logbase    1.0001      1.000100e+00
-logfn              
-logspec    no      no
-lowerf     133.33334   1.333333e+02
-lpbeam     1e-40       1.000000e-40
-lponlybeam 7e-29       7.000000e-29
-lw     6.5     6.500000e+00
-maxhmmpf   -1      -1
-maxnewoov  20      20
-maxwpf     -1      -1
-mdef               
-mean               
-mfclogdir          
-min_endfr  0       0
-mixw               
-mixwfloor  0.0000001   1.000000e-07
-mllr               
-mmap       yes     yes
-ncep       13      13
-nfft       512     512
-nfilt      40      40
-nwpen      1.0     1.000000e+00
-pbeam      1e-48       1.000000e-48
-pip        1.0     1.000000e+00
-pl_beam    1e-10       1.000000e-10
-pl_pbeam   1e-5        1.000000e-05
-pl_window  0       0
-rawlogdir          
-remove_dc  no      no
-round_filters  yes     yes
-samprate   16000       1.600000e+04
-seed       -1      -1
-sendump            
-senlogdir          
-senmgau            
-silprob    0.005       5.000000e-03
-smoothspec no      no
-svspec             
-tmat               
-tmatfloor  0.0001      1.000000e-04
-topn       4       4
-topn_beam  0       0
-toprule            
-transform  legacy      legacy
-unit_area  yes     yes
-upperf     6855.4976   6.855498e+03
-usewdphones    no      no
-uw     1.0     1.000000e+00
-var                
-varfloor   0.0001      1.000000e-04
-varnorm    no      no
-verbose    no      no
-warp_params            
-warp_type  inverse_linear  inverse_linear
-wbeam      7e-29       7.000000e-29
-wip        0.65        6.500000e-01
-wlen       0.025625    2.562500e-02

INFO: cmd_ln.c(691): Parsing command line:
\
    -nfilt 20 \
    -lowerf 1 \
    -upperf 4000 \
    -wlen 0.025 \
    -transform dct \
    -round_filters no \
    -remove_dc yes \
    -svspec 0-12/13-25/26-38 \
    -feat 1s_c_d_dd \
    -agc none \
    -cmn current \
    -cmninit 56,-3,1 \
    -varnorm no

Current configuration:
[NAME]      [DEFLT]     [VALUE]
-agc        none        none
-agcthresh  2.0     2.000000e+00
-alpha      0.97        9.700000e-01
-ceplen     13      13
-cmn        current     current
-cmninit    8.0     56,-3,1
-dither     no      no
-doublebw   no      no
-feat       1s_c_d_dd   1s_c_d_dd
-frate      100     100
-input_endian   little      little
-lda                
-ldadim     0       0
-lifter     0       0
-logspec    no      no
-lowerf     133.33334   1.000000e+00
-ncep       13      13
-nfft       512     512
-nfilt      40      20
-remove_dc  no      yes
-round_filters  yes     no
-samprate   16000       8.000000e+03
-seed       -1      -1
-smoothspec no      no
-svspec             0-12/13-25/26-38
-transform  legacy      dct
-unit_area  yes     yes
-upperf     6855.4976   4.000000e+03
-varnorm    no      no
-verbose    no      no
-warp_params            
-warp_type  inverse_linear  inverse_linear
-wlen       0.025625    2.500000e-02

INFO: acmod.c(242): Parsed model-specific feature parameters from /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/feat.params
INFO: feat.c(684): Initializing feature stream to type: '1s_c_d_dd', ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean[0]= 12.00, mean[1..12]= 0.0
INFO: acmod.c(163): Using subvector specification 0-12/13-25/26-38
INFO: mdef.c(520): Reading model definition: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/mdef
INFO: mdef.c(531): Found byte-order mark BMDF, assuming this is a binary mdef file
INFO: bin_mdef.c(330): Reading binary model definition: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/mdef
INFO: bin_mdef.c(507): 50 CI-phone, 143047 CD-phone, 3 emitstate/phone, 150 CI-sen, 5150 Sen, 27135 Sen-Seq
INFO: tmat.c(205): Reading HMM transition probability matrices: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/transition_matrices
INFO: acmod.c(117): Attempting to use SCHMM computation module
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/means
INFO: ms_gauden.c(292): 1 codebook, 3 feature, size: 
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/variances
INFO: ms_gauden.c(292): 1 codebook, 3 feature, size: 
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(294):  256x13
INFO: ms_gauden.c(354): 0 variance values floored
INFO: s2_semi_mgau.c(908): Loading senones from dump file /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/sendump
INFO: s2_semi_mgau.c(932): BEGIN FILE FORMAT DESCRIPTION
INFO: s2_semi_mgau.c(1027): Using memory-mapped I/O for senones
INFO: s2_semi_mgau.c(1304): Maximum top-N: 4 Top-N beams: 0 0 0
INFO: dict.c(306): Allocating 137127 * 20 bytes (2678 KiB) for word entries
INFO: dict.c(321): Reading main dictionary: /mnt/sdcard/speechbrowser/temp.dict
INFO: dict.c(212): Allocated 1004 KiB for strings, 1657 KiB for phones
INFO: dict.c(324): 133021 words read
INFO: dict.c(330): Reading filler dictionary: /mnt/sdcard/speechbrowser/speech/hmm/en_US/hub4wsj_sc_8k/noisedict
INFO: dict.c(212): Allocated 0 KiB for strings, 0 KiB for phones
INFO: dict.c(333): 11 words read
INFO: dict2pid.c(396): Building PID tables for dictionary
INFO: dict2pid.c(404): Allocating 50^3 * 2 bytes (244 KiB) for word-initial triphones
INFO: dict2pid.c(131): Allocated 30200 bytes (29 KiB) for word-final triphones
INFO: dict2pid.c(195): Allocated 30200 bytes (29 KiB) for single-phone word triphones
INFO: fsg_search.c(145): FSG(beam: -1080, pbeam: -1080, wbeam: -634; wip: -26, pip: 0)
INFO: jsgf.c(546): Defined rule: PUBLIC <test.hi>
INFO: fsg_model.c(213): Computing transitive closure for null transitions
INFO: fsg_model.c(264): 0 null transitions added
INFO: fsg_model.c(411): Adding silence transitions for <sil> to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++NOISE++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++BREATH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++SMACK++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++COUGH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++LAUGH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++TONE++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++UH++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_model.c(411): Adding silence transitions for ++UM++ to FSG
INFO: fsg_model.c(431): Added 53 silence word transitions
INFO: fsg_search.c(364): Added 13 alternate word transitions
INFO: fsg_lextree.c(108): Allocated 5406 bytes (5 KiB) for left and right context phones
INFO: fsg_lextree.c(251): 715 HMM nodes in lextree (586 leaves)
INFO: fsg_lextree.c(253): Allocated 77220 bytes (75 KiB) for all lextree nodes
INFO: fsg_lextree.c(256): Allocated 63288 bytes (61 KiB) for lextree leafnodes

    Thanks

     

  • Nickolay V. Shmyrev

    Nickolay V. Shmyrev - 2012-08-08

    The audio has zero energy regions. You need to add "-dither yes" to engine
    configuration.

     

  • Jon

    Jon - 2012-08-08

    That got it! Thank you, this problem has been dogging me for a while, and I
    couldn't for the life of me figure it out.

    Is there some documentation that you could point me to which elaborates on the
    -dither option (and other command line options for that matter)?

    Once again, thank you.

     

  • Nickolay V. Shmyrev

    Nickolay V. Shmyrev - 2012-08-09

    Is there some documentation that you could point me to which elaborates on
    the -dither option (and other command line options for that matter)?

    man pocketsphinx_batch

     

Log in to post a comment.