Menu ▾ ▴

Missing word "six"

Help
Chris C. Russo
2016-03-21
2016-03-25

  • Chris C. Russo

    Chris C. Russo - 2016-03-21

    Hello CMU Sphinx users!

    I've been working with the stand-alone pre-compiled version of pocketsphinx for the last 48 hours. I found the technology really interesting and I'm just practicing a little bit. At this point, I've a reduced dictionary, as the only thing I have to detect are numbers.

    However for some reason, it's actually properly detecting all the numbers, but not the "six".

    Command I'm running:
    ./pocketsphinx_continuous.exe -infile files/01_16.wav -hmm model/en-us/en-us -lm model/en-us/en-us.lm.bin -dict model/en-us/micro.dict -logfn record.txt -lw 1

    The micro.dict file

    zero Z IH R OW
    zero(2) Z IY R OW
    one W AH N
    one(2) HH W AH N
    two T UW
    three TH R IY
    four F AO R
    five F AY V
    six S IH T S
    six(2) S IH K S IH Z
    six(3) S IH K S T
    seven S EH V AH N
    eight EY T
    nine N AY N

    Verbose / logs:

    $ ./pocketsphinx_continuous.exe -infile files/01_16.wav -hmm model/en-us/en-us -lm model/en-us/en-us.lm.bin -dict model/en-us/micro.dict -lw 1 -debug 2

    INFO: pocketsphinx.c(152): Parsed model-specific feature parameters from model/en-us/en-us/feat.params
    Current configuration:
    [NAME] [DEFLT] [VALUE]
    -agc none none
    -agcthresh 2.0 2.000000e+000
    -allphone
    -allphone_ci no no
    -alpha 0.97 9.700000e-001
    -ascale 20.0 2.000000e+001
    -aw 1 1
    -backtrace no no
    -beam 1e-48 1.000000e-048
    -bestpath yes yes
    -bestpathlw 9.5 9.500000e+000
    -ceplen 13 13
    -cmn current current
    -cmninit 8.0 40,3,-1
    -compallsen no no
    -debug 2
    -dict model/en-us/micro.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-008
    -frate 100 100
    -fsg
    -fsgusealtpron yes yes
    -fsgusefiller yes yes
    -fwdflat yes yes
    -fwdflatbeam 1e-64 1.000000e-064
    -fwdflatefwid 4 4
    -fwdflatlw 8.5 8.500000e+000
    -fwdflatsfwin 25 25
    -fwdflatwbeam 7e-29 7.000000e-029
    -fwdtree yes yes
    -hmm model/en-us/en-us
    -input_endian little little
    -jsgf
    -keyphrase
    -kws
    -kws_delay 10 10
    -kws_plp 1e-1 1.000000e-001
    -kws_threshold 1 1.000000e+000
    -latsize 5000 5000
    -lda
    -ldadim 0 0
    -lifter 0 22
    -lm model/en-us/en-us.lm.bin
    -lmctl
    -lmname
    -logbase 1.0001 1.000100e+000
    -logfn
    -logspec no no
    -lowerf 133.33334 1.300000e+002
    -lpbeam 1e-40 1.000000e-040
    -lponlybeam 7e-29 7.000000e-029
    -lw 6.5 1.000000e+000
    -maxhmmpf 30000 30000
    -maxwpf -1 -1
    -mdef
    -mean
    -mfclogdir
    -min_endfr 0 0
    -mixw
    -mixwfloor 0.0000001 1.000000e-007
    -mllr
    -mmap yes yes
    -ncep 13 13
    -nfft 512 512
    -nfilt 40 25
    -nwpen 1.0 1.000000e+000
    -pbeam 1e-48 1.000000e-048
    -pip 1.0 1.000000e+000
    -pl_beam 1e-10 1.000000e-010
    -pl_pbeam 1e-10 1.000000e-010
    -pl_pip 1.0 1.000000e+000
    -pl_weight 3.0 3.000000e+000
    -pl_window 5 5
    -rawlogdir
    -remove_dc no no
    -remove_noise yes yes
    -remove_silence yes yes
    -round_filters yes yes
    -samprate 16000 1.600000e+004
    -seed -1 -1
    -sendump
    -senlogdir
    -senmgau
    -silprob 0.005 5.000000e-003
    -smoothspec no no
    -svspec
    -tmat
    -tmatfloor 0.0001 1.000000e-004
    -topn 4 4
    -topn_beam 0 0
    -toprule
    -transform legacy dct
    -unit_area yes yes
    -upperf 6855.4976 6.800000e+003
    -uw 1.0 1.000000e+000
    -vad_postspeech 50 50
    -vad_prespeech 20 20
    -vad_startspeech 10 10
    -vad_threshold 2.0 2.000000e+000
    -var
    -varfloor 0.0001 1.000000e-004
    -varnorm no no
    -verbose no no
    -warp_params
    -warp_type inverse_linear inverse_linear
    -wbeam 7e-29 7.000000e-029
    -wip 0.65 6.500000e-001
    -wlen 0.025625 2.562500e-002

    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: acmod.c(154): Reading linear feature transformation from model/en-us/en-us/feature_transform
    INFO: mdef.c(518): Reading model definition: model/en-us/en-us/mdef
    INFO: bin_mdef.c(181): Allocating 142124 * 8 bytes (1110 KiB) for CD tree
    INFO: tmat.c(206): Reading HMM transition probability matrices: model/en-us/en-us/transition_matrices
    INFO: acmod.c(117): Attempting to use PTM computation module
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/means
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/variances
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(354): 813 variance values floored
    INFO: ptm_mgau.c(801): Number of codebooks exceeds 256: 5138
    INFO: acmod.c(119): Attempting to use semi-continuous computation module
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/means
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/variances
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(354): 813 variance values floored
    INFO: acmod.c(121): Falling back to general multi-stream GMM computation
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/means
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(198): Reading mixture gaussian parameter: model/en-us/en-us/variances
    INFO: ms_gauden.c(292): 5138 codebook, 1 feature, size:
    INFO: ms_gauden.c(294): 32x36
    INFO: ms_gauden.c(354): 813 variance values floored
    INFO: ms_senone.c(149): Reading senone mixture weights: model/en-us/en-us/mixture_weights
    INFO: ms_senone.c(200): Truncating senone logs3(pdf) values by 10 bits
    INFO: ms_senone.c(207): Not transposing mixture weights in memory
    INFO: ms_senone.c(268): Read mixture weights for 5138 senones: 1 features x 32 codewords
    INFO: ms_senone.c(320): Mapping senones to individual codebooks
    INFO: ms_mgau.c(141): The value of topn: 4
    INFO: phone_loop_search.c(114): State beam -225 Phone exit beam -225 Insertion penalty 0
    INFO: dict.c(320): Allocating 4119 * 32 bytes (128 KiB) for word entries
    INFO: dict.c(333): Reading main dictionary: model/en-us/micro.dict
    INFO: dict.c(213): Allocated 0 KiB for strings, 0 KiB for phones
    INFO: dict.c(336): 14 words read
    INFO: dict.c(358): Reading filler dictionary: model/en-us/en-us/noisedict
    INFO: dict.c(213): Allocated 0 KiB for strings, 0 KiB for phones
    INFO: dict.c(361): 9 words read
    INFO: dict2pid.c(396): Building PID tables for dictionary
    INFO: dict2pid.c(406): Allocating 46^3 * 2 bytes (190 KiB) for word-initial triphones
    INFO: dict2pid.c(132): Allocated 51152 bytes (49 KiB) for word-final triphones
    INFO: dict2pid.c(196): Allocated 51152 bytes (49 KiB) for single-phone word triphones
    INFO: ngram_model_trie.c(347): Trying to read LM in trie binary format
    INFO: ngram_search_fwdtree.c(99): 12 unique initial diphones
    INFO: ngram_search_fwdtree.c(148): 0 root, 0 non-root channels, 10 single-phone words
    INFO: ngram_search_fwdtree.c(186): Creating search tree
    INFO: ngram_search_fwdtree.c(192): before: 0 root, 0 non-root channels, 10 single-phone words
    INFO: ngram_search_fwdtree.c(326): after: max nonroot chan increased to 149
    INFO: ngram_search_fwdtree.c(339): after: 12 root, 21 non-root channels, 9 single-phone words
    INFO: ngram_search_fwdflat.c(157): fwdflat: min_ef_width = 4, max_sf_win = 25
    INFO: continuous.c(307): C:\Users\admin\Desktop\modelo1\bin\Release\x64\pocketsphinx_continuous.exe COMPILED ON: Jan 24 2016, AT: 07:35:22

    INFO: cmn_prior.c(131): cmn_prior_update: from < 40.00 3.00 -1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 59.77 25.73 -30.09 31.64 -17.87 -3.09 5.67 -16.70 17.23 -13.99 1.21 5.62 -9.56 >
    INFO: ngram_search_fwdtree.c(1553): 1615 words recognized (6/fr)
    INFO: ngram_search_fwdtree.c(1555): 45610 senones evaluated (170/fr)
    INFO: ngram_search_fwdtree.c(1559): 27981 channels searched (104/fr), 2503 1st, 23331 last
    INFO: ngram_search_fwdtree.c(1562): 2207 words for which last channels evaluated (8/fr)
    INFO: ngram_search_fwdtree.c(1564): 926 candidate words for entering last phone (3/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.16 CPU 0.058 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.14 wall 0.054 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 10 words
    INFO: ngram_search_fwdflat.c(948): 859 words recognized (3/fr)
    INFO: ngram_search_fwdflat.c(950): 35278 senones evaluated (131/fr)
    INFO: ngram_search_fwdflat.c(952): 20950 channels searched (77/fr)
    INFO: ngram_search_fwdflat.c(954): 2432 words searched (9/fr)
    INFO: ngram_search_fwdflat.c(957): 1039 word transitions (3/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.05 CPU 0.017 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.06 wall 0.020 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .233
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 98 nodes, 7 links
    INFO: ps_lattice.c(1380): Besteight two
    eight two two one
    path score: -9352
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:233:267) = -612077
    INFO: ps_lattice.c(1441): Joint P(O,S) = -619201 P(S|O) = -7124
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 59.77 25.73 -30.09 31.64 -17.87 -3.09 5.67 -16.70 17.23 -13.99 1.21 5.62 -9.56 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 57.05 24.79 -29.11 34.55 -18.54 -2.06 6.49 -20.19 14.28 -14.35 -0.42 5.39 -8.80 >
    INFO: ngram_search_fwdtree.c(1553): 3021 words recognized (9/fr)
    INFO: ngram_search_fwdtree.c(1555): 83996 senones evaluated (246/fr)
    INFO: ngram_search_fwdtree.c(1559): 52714 channels searched (154/fr), 3981 1st, 44159 last
    INFO: ngram_search_fwdtree.c(1562): 4008 words for which last channels evaluated (11/fr)
    INFO: ngram_search_fwdtree.c(1564): 1633 candidate words for entering last phone (4/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.23 CPU 0.068 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.24 wall 0.070 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 15 words
    INFO: ngram_search_fwdflat.c(948): 2900 words recognized (8/fr)
    INFO: ngram_search_fwdflat.c(950): 98768 senones evaluated (289/fr)
    INFO: ngram_search_fwdflat.c(952): 70226 channels searched (205/fr)
    INFO: ngram_search_fwdflat.c(954): 5789 words searched (16/fr)
    INFO: ngram_search_fwdflat.c(957): 2346 word transitions (6/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.19 CPU 0.055 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.18 wall 0.053 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .299
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 578 nodes, 2651 links
    INFO: ps_lattice.c(1380): Bestpath score: -15225
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:299:340) = -833498
    INFO: ps_lattice.c(1441): Joint P(O,S) = -973462 P(S|O) = -139964
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 57.05 24.79 -29.11 34.55 -18.54 -2.06 6.49 -20.19 14.28 -14.35 -0.42 5.39 -8.80 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 57.35 25.83 -31.31 37.35 -18.26 -1.94 5.26 -21.96 14.76 -14.39 0.19 5.88 -8.60 >
    INFO: cmn_prior.c(99): cmn_prior_update: from < 57.35 25.83 -31.31 37.35 -18.26 -1.94 5.26 -21.96 14.76 -14.39 0.19 5.88 -8.60 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 56.83 29.95 -32.22 36.91 -19.88 -0.93 6.56 -24.45 14.15 -15.08 -0.21 6.80 -8.44 >
    INFO: ngram_search.c(459): Resized backpointer table to 10000 entries
    INFO: cmn_prior.c(131): cmn_prior_update: from < 56.83 29.95 -32.22 36.91 -19.88 -0.93 6.56 -24.45 14.15 -15.08 -0.21 6.80 -8.44 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 52.44 27.33 -34.54 34.08 -19.19 -1.27 4.84 -24.13 13.82 -14.10 -0.18 7.15 -8.00 >
    INFO: ngram_search_fwdtree.c(1553): 5418 words recognized (8/fr)
    INFO: ngram_search_fwdtree.c(1555): 161043 senones evaluated (229/fr)
    INFO: ngram_search_fwdtree.c(1559): 100628 channels searched (142/fr), 8192 1st, 83757 last
    INFO: ngram_search_fwdtree.c(1562): 7635 words for which last channels evaluated (10/fr)
    INFO: ngram_search_fwdtree.c(1564): 3047 candidate words for entering last phone (4/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.47 CPU 0.066 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.47 wall 0.066 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 14 words
    INFO: ngram_search_fwdflat.c(948): 5481 words recognized (8/fr)
    INFO: ngram_search_fwdflat.c(950): 203107 senones evaluated (289/fr)
    INFO: ngram_search_fwdflat.c(952): 145265 channels searched (206/fr)
    INFO: ngram_search_fwdflat.c(954): 12049 words searched (17/fr)
    INFO: ngram_search_fwdflat.c(957): 5062 word transitions (7/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.37 CPU 0.053 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.36 wall 0.052 xRT
    INFO: ngram_search.nine three one two one two one three
    three
    c(1253): lattice start node .0 end node .640
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 1141 nodes, 4596 links
    INFO: ps_lattice.c(1380): Bestpath score: -34289
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:640:702) = -1840914
    INFO: ps_lattice.c(1441): Joint P(O,S) = -2111102 P(S|O) = -270188
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.01 wall 0.001 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 52.44 27.33 -34.54 34.08 -19.19 -1.27 4.84 -24.13 13.82 -14.10 -0.18 7.15 -8.00 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 52.28 28.18 -34.33 35.30 -19.53 -1.45 4.96 -25.44 13.61 -14.04 0.13 7.98 -7.48 >
    INFO: ngram_search_fwdtree.c(1553): 1294 words recognized (9/fr)
    INFO: ngram_search_fwdtree.c(1555): 32716 senones evaluated (237/fr)
    INFO: ngram_search_fwdtree.c(1559): 20092 channels searched (145/fr), 1608 1st, 16562 last
    INFO: ngram_search_fwdtree.c(1562): 1628 words for which last channels evaluated (11/fr)
    INFO: ngram_search_fwdtree.c(1564): 626 candidate words for entering last phone (4/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.09 CPU 0.068 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.09 wall 0.069 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 13 words
    INFO: ngram_search_fwdflat.c(948): 1101 words recognized (8/fr)
    INFO: ngram_search_fwdflat.c(950): 40711 senones evaluated (295/fr)
    INFO: ngram_search_fwdflat.c(952): 29155 channels searched (211/fr)
    INFO: ngram_search_fwdflat.c(954): 2406 words searched (17/fr)
    INFO: ngram_search_fwdflat.c(957): 899 word transitions (6/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.08 CPU 0.057 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.08 wall 0.054 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .121
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 242 nodes, 1043 links
    INFO: ps_lattice.c(1380): Bestpath score: -6526
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:121:136) = -318823
    INFO: ps_lattice.c(1441): Joint P(O,S) = -402781 P(S|O) = -83958
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 52.28 28.18 -34.33 35.30 -19.53 -1.45 4.96 -25.44 13.61 -14.04 0.13 7.98 -7.48 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 52.24 28.18 -34.25 35.34 -19.45 -1.51 4.92 -25.40 13.60 -14.09 0.05 8.05 -7.44 >
    INFO: cmn_prior.c(131): cmn_prior_update: from < 52.24 28.18 -34.25 35.34 -19.45 -1.51 4.92 -25.40 13.60 -14.09 0.05 8.05 -7.44 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 49.34 25.26 -34.22 34.17 -19.44 -2.38 3.70 -23.77 12.56 -13.28 -0.27 7.69 -7.60 >
    INFO: ngram_search_fwdtree.c(1553): 578 words recognized (5/fr)
    INFO: ngram_search_fwdtree.c(1555): 16717 senones evaluated (148/fr)
    INFO: ngram_search_fwdtree.c(1559): 8290 channels searched (73/fr), 1308 1st, 5881 last
    INFO: ngram_search_fwdtree.c(1562): 935 words for which last channels evaluated (8/fr)
    INFO: ngram_search_fwdtree.c(1564): 178 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.06 CPU 0.055 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.05 wall 0.049 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 6 words
    INFO: ngram_search_fwdflat.c(948): 652 words recognized (6/fr)
    INFO: ngram_search_fwdflat.c(950): 15326 senones evaluated (136/fr)
    INFO: ngram_search_fwdflat.c(952): 7913 channels searched (70/fr)
    INFO: ngram_search_fwdflat.c(954): 1313 words searched (11/fr)
    INFO: ngram_search_fwdflat.c(957): 328 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.014 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.03 wall 0.022 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .70
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_seaeight
    three
    eight
    rch.c(1384): Lattice has 182 nodes, 688 links
    INFO: ps_lattice.c(1380): Bestpath score: -2709
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:70:111) = -158830
    INFO: ps_lattice.c(1441): Joint P(O,S) = -217931 P(S|O) = -59101
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 49.34 25.26 -34.22 34.17 -19.44 -2.38 3.70 -23.77 12.56 -13.28 -0.27 7.69 -7.60 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 48.91 24.86 -33.06 36.03 -18.79 -2.21 3.76 -24.92 11.76 -13.39 -0.86 7.67 -7.61 >
    INFO: ngram_search_fwdtree.c(1553): 369 words recognized (4/fr)
    INFO: ngram_search_fwdtree.c(1555): 12360 senones evaluated (129/fr)
    INFO: ngram_search_fwdtree.c(1559): 6138 channels searched (63/fr), 942 1st, 4592 last
    INFO: ngram_search_fwdtree.c(1562): 723 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 132 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.049 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.04 wall 0.047 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 5 words
    INFO: ngram_search_fwdflat.c(948): 348 words recognized (4/fr)
    INFO: ngram_search_fwdflat.c(950): 11123 senones evaluated (116/fr)
    INFO: ngram_search_fwdflat.c(952): 5789 channels searched (60/fr)
    INFO: ngram_search_fwdflat.c(954): 1051 words searched (10/fr)
    INFO: ngram_search_fwdflat.c(957): 192 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.01 wall 0.016 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .71
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 73 nodes, 24 links
    INFO: ps_lattice.c(1380): Bestpath score: -2982
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:71:94) = -216221
    INFO: ps_lattice.c(1441): Joint P(O,S) = -222560 P(S|O) = -6339
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 48.91 24.86 -33.06 36.03 -18.79 -2.21 3.76 -24.92 11.76 -13.39 -0.86 7.67 -7.61 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 46.48 22.58 -31.30 35.09 -18.85 -2.14 3.31 -23.77 10.67 -12.62 -1.45 7.33 -7.58 >
    INFO: ngram_search_fwdtree.c(1553): 340 words recognized (3/fr)
    INFO: ngram_search_fwdtree.c(1555): 11470 senones evaluated (117/fr)
    INFO: ngram_search_fwdtree.c(1559): 5543 channels searched (56/fr), 1128 1st, 3736 last
    INFO: ngram_search_fwdtree.c(1562): 745 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 139 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.048 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.04 wall 0.041 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 4 words
    INFO: ngram_search_fwdflat.c(948): 403 words recognized (4/fr)
    INFO: ngram_search_fwdflat.c(950): 8249 senones evaluated (84/fr)
    INFO: ngram_search_fwdflat.c(952): 3816 channels searched (38/fr)
    INFO: ngram_search_fwdflat.c(954): 991 words searched (10/fr)
    INFO: ngram_search_fwdflat.c(957): 196 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.02 wall 0.015 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .41
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 59 nodes, 8 links
    INFO: ps_lattice.c(1380): Bestpath score: -1566
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:41:96) = -177960
    INFO: ps_lattice.c(1441): Joint P(O,S) = -183524 P(S|O) = -5564
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 46.48 22.58 -31.30 35.09 -18.8three
    four
    5 -2.14 3.31 -23.77 10.67 -12.62 -1.45 7.33 -7.58 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 46.94 23.01 -30.93 35.85 -18.36 -2.42 3.12 -24.38 10.31 -12.80 -1.62 7.28 -7.63 >
    INFO: cmn_prior.c(131): cmn_prior_update: from < 46.94 23.01 -30.93 35.85 -18.36 -2.42 3.12 -24.38 10.31 -12.80 -1.62 7.28 -7.63 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 45.85 22.21 -30.51 36.61 -18.34 -1.92 3.50 -24.75 10.08 -12.76 -2.12 7.26 -7.53 >
    INFO: ngram_search_fwdtree.c(1553): 443 words recognized (5/fr)
    INFO: ngram_search_fwdtree.c(1555): 14118 senones evaluated (147/fr)
    INFO: ngram_search_fwdtree.c(1559): 7329 channels searched (76/fr), 1033 1st, 5480 last
    INFO: ngram_search_fwdtree.c(1562): 803 words for which last channels evaluated (8/fr)
    INFO: ngram_search_fwdtree.c(1564): 207 candidate words for entering last phone (2/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.049 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.05 wall 0.052 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 7 words
    INFO: ngram_search_fwdflat.c(948): 448 words recognized (5/fr)
    INFO: ngram_search_fwdflat.c(950): 14415 senones evaluated (150/fr)
    INFO: ngram_search_fwdflat.c(952): 7999 channels searched (83/fr)
    INFO: ngram_search_fwdflat.c(954): 1162 words searched (12/fr)
    INFO: ngram_search_fwdflat.c(957): 276 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.03 wall 0.026 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .70
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 97 nodes, 85 links
    INFO: ps_lattice.c(1380): Bestpath score: -3291
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:70:94) = -234978
    INFO: ps_lattice.c(1441): Joint P(O,S) = -246879 P(S|O) = -11901
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 45.85 22.21 -30.51 36.61 -18.34 -1.92 3.50 -24.75 10.08 -12.76 -2.12 7.26 -7.53 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 44.59 22.35 -29.73 33.42 -18.83 -1.85 3.43 -23.40 10.23 -12.11 -1.93 6.88 -7.91 >
    INFO: ngram_search_fwdtree.c(1553): 384 words recognized (4/fr)
    INFO: ngram_search_fwdtree.c(1555): 13300 senones evaluated (130/fr)
    INFO: ngram_search_fwdtree.c(1559): 6294 channels searched (61/fr), 1075 1st, 4590 last
    INFO: ngram_search_fwdtree.c(1562): 741 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 161 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.046 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.04 wall 0.044 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 6 words
    INFO: ngram_search_fwdflat.c(948): 418 words recognized (4/fr)
    INFO: ngram_search_fwdflat.c(950): 12487 senones evaluated (122/fr)
    INFO: ngram_search_fwdflat.c(952): 5910 channels searched (57/fr)
    INFO: ngram_search_fwdflat.c(954): 1100 words searched (10/fr)
    INFO: ngram_search_fwdflat.c(957): 246 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.03 CPU 0.031 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.02 wall 0.020 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .63
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 115 nodes, 264 links
    INFO: ps_lattice.c(1380): Bestpath score: -2326
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:63:100) = -155842
    INFO: ps_lattice.c(1441): Joint P(O,S) = -176621 P(S|O) = -20779
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 44.59 22.35 -29.73 33.42 -18.83 -1.85 3.43 -23.40 10.23 -12.11 -1.93 6.88 -7.91 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 42.50 20.30 -28.31 32.41 -18.73 -2.11 3.00 eight
    eight
    -22.53 9.16 -11.35 -2.47 6.63 -7.97 >
    INFO: ngram_search_fwdtree.c(1553): 342 words recognized (3/fr)
    INFO: ngram_search_fwdtree.c(1555): 10871 senones evaluated (111/fr)
    INFO: ngram_search_fwdtree.c(1559): 5153 channels searched (52/fr), 1120 1st, 3290 last
    INFO: ngram_search_fwdtree.c(1562): 731 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 121 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.03 CPU 0.032 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.04 wall 0.041 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 4 words
    INFO: ngram_search_fwdflat.c(948): 407 words recognized (4/fr)
    INFO: ngram_search_fwdflat.c(950): 8674 senones evaluated (89/fr)
    INFO: ngram_search_fwdflat.c(952): 4033 channels searched (41/fr)
    INFO: ngram_search_fwdflat.c(954): 994 words searched (10/fr)
    INFO: ngram_search_fwdflat.c(957): 197 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.01 wall 0.015 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .41
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 58 nodes, 10 links
    INFO: ps_lattice.c(1380): Bestpath score: -1667
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:41:96) = -178631
    INFO: ps_lattice.c(1441): Joint P(O,S) = -186341 P(S|O) = -7710
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 42.50 20.30 -28.31 32.41 -18.73 -2.11 3.00 -22.53 9.16 -11.35 -2.47 6.63 -7.97 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 43.99 21.56 -29.15 33.13 -18.80 -2.60 3.12 -22.33 10.44 -11.57 -2.49 6.94 -8.30 >
    INFO: cmn_prior.c(131): cmn_prior_update: from < 43.99 21.56 -29.15 33.13 -18.80 -2.60 3.12 -22.33 10.44 -11.57 -2.49 6.94 -8.30 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 44.32 22.38 -28.86 34.02 -18.96 -2.53 3.59 -23.32 9.57 -11.50 -2.32 7.94 -7.24 >
    INFO: ngram_search_fwdtree.c(1553): 1814 words recognized (8/fr)
    INFO: ngram_search_fwdtree.c(1555): 50657 senones evaluated (237/fr)
    INFO: ngram_search_fwdtree.c(1559): 33753 channels searched (157/fr), 2508 1st, 28418 last
    INFO: ngram_search_fwdtree.c(1562): 2472 words for which last channels evaluated (11/fr)
    INFO: ngram_search_fwdtree.c(1564): 1040 candidate words for entering last phone (4/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.14 CPU 0.066 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.15 wall 0.068 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 13 words
    INFO: ngram_search_fwdflat.c(948): 1716 words recognized (8/fr)
    INFO: ngram_search_fwdflat.c(950): 53713 senones evaluated (251/fr)
    INFO: ngram_search_fwdflat.c(952): 40212 channels searched (187/fr)
    INFO: ngram_search_fwdflat.c(954): 3563 words searched (16/fr)
    INFO: ngram_search_fwdflat.c(957): 1289 word transitions (6/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.09 CPU 0.044 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.10 wall 0.044 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .180
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 325 nodes, 945 links
    INFO: ps_lattice.c(1380): Bestpath score: -10136
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:180:212) = -541776
    INFO: ps_lattice.c(1441): Joint P(O,S) = -597424 P(S|O) = -55648
    INFO: ngram_search.c(875): bestpath 0.02 CPU 0.007 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.002 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 44.32 22.38 -28.86 34.02 -18.96 -2.53 3.59 -23.32 9.57 -11.50 -2.32 7.94 -7.24 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 42.66 20.57 -29.23 32.94 -18.92 -3.13 2.74 -22.33 9.12 -11.14 -2.20 7.91 -7.24 >
    INFO: ngram_search_fwdtree.c(1553): 709 words recognized (6/fr)
    INFO: ngram_search_fwdtree.c(1555eight
    three
    ): 18161 senones evaluated (161/fr)
    INFO: ngram_search_fwdtree.c(1559): 9526 channels searched (84/fr), 1308 1st, 6976 last
    INFO: ngram_search_fwdtree.c(1562): 1051 words for which last channels evaluated (9/fr)
    INFO: ngram_search_fwdtree.c(1564): 238 candidate words for entering last phone (2/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.041 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.06 wall 0.049 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 9 words
    INFO: ngram_search_fwdflat.c(948): 699 words recognized (6/fr)
    INFO: ngram_search_fwdflat.c(950): 19802 senones evaluated (175/fr)
    INFO: ngram_search_fwdflat.c(952): 11755 channels searched (104/fr)
    INFO: ngram_search_fwdflat.c(954): 1563 words searched (13/fr)
    INFO: ngram_search_fwdflat.c(957): 461 word transitions (4/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.05 CPU 0.041 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.03 wall 0.031 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .71
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 215 nodes, 653 links
    INFO: ps_lattice.c(1380): Bestpath score: -3081
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:71:111) = -147851
    INFO: ps_lattice.c(1441): Joint P(O,S) = -219876 P(S|O) = -72025
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 42.66 20.57 -29.23 32.94 -18.92 -3.13 2.74 -22.33 9.12 -11.14 -2.20 7.91 -7.24 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 43.14 20.86 -28.69 34.88 -18.52 -2.94 2.83 -23.46 8.86 -11.52 -2.61 7.87 -7.28 >
    INFO: ngram_search_fwdtree.c(1553): 540 words recognized (6/fr)
    INFO: ngram_search_fwdtree.c(1555): 16354 senones evaluated (172/fr)
    INFO: ngram_search_fwdtree.c(1559): 9447 channels searched (99/fr), 991 1st, 7517 last
    INFO: ngram_search_fwdtree.c(1562): 841 words for which last channels evaluated (8/fr)
    INFO: ngram_search_fwdtree.c(1564): 258 candidate words for entering last phone (2/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.049 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.06 wall 0.058 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 11 words
    INFO: ngram_search_fwdflat.c(948): 503 words recognized (5/fr)
    INFO: ngram_search_fwdflat.c(950): 18180 senones evaluated (191/fr)
    INFO: ngram_search_fwdflat.c(952): 10849 channels searched (114/fr)
    INFO: ngram_search_fwdflat.c(954): 1314 words searched (13/fr)
    INFO: ngram_search_fwdflat.c(957): 419 word transitions (4/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.03 CPU 0.033 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.03 wall 0.032 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .72
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 127 nodes, 240 links
    INFO: ps_lattice.c(1380): Bestpath score: -3670
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:72:93) = -232142
    INFO: ps_lattice.c(1441): Joint P(O,S) = -259833 P(S|O) = -27691
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 43.14 20.86 -28.69 34.88 -18.52 -2.94 2.83 -23.46 8.86 -11.52 -2.61 7.87 -7.28 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 42.94 20.59 -28.46 34.80 -18.73 -3.08 2.81 -23.33 8.81 -11.40 -2.58 7.76 -7.23 >
    INFO: cmn_prior.c(131): cmn_prior_update: from < 42.94 20.59 -28.46 34.80 -18.73 -3.08 2.81 -23.33 8.81 -11.40 -2.58 7.76 -7.23 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 40.78 18.59 -27.18 33.79 -18.68 -2.83 2.36 -22.40 7.99 -10.87 -3.14 7.41 -7.32 >
    INFO: ngram_search_fwdtree.c(1553): 352 words recognized (4/fr)
    INFO: ngram_search_fwdtree.c(1555): 11636 senones evaluated (119/fr)
    INFO: ngram_search_fwdtree.c(1559): 5600 channels searched (57/fr), 1128 1st, 3717 last
    INFO: ngram_seight
    three
    earch_fwdtree.c(1562): 753 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 145 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.048 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.04 wall 0.046 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 5 words
    INFO: ngram_search_fwdflat.c(948): 446 words recognized (5/fr)
    INFO: ngram_search_fwdflat.c(950): 10728 senones evaluated (109/fr)
    INFO: ngram_search_fwdflat.c(952): 5171 channels searched (52/fr)
    INFO: ngram_search_fwdflat.c(954): 1080 words searched (11/fr)
    INFO: ngram_search_fwdflat.c(957): 263 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.01 wall 0.015 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .54
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 85 nodes, 66 links
    INFO: ps_lattice.c(1380): Bestpath score: -2391
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:54:96) = -174789
    INFO: ps_lattice.c(1441): Joint P(O,S) = -194174 P(S|O) = -19385
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 40.78 18.59 -27.18 33.79 -18.68 -2.83 2.36 -22.40 7.99 -10.87 -3.14 7.41 -7.32 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 41.35 18.96 -26.87 35.72 -18.15 -2.71 2.54 -23.72 7.68 -11.25 -3.48 7.39 -7.41 >
    INFO: ngram_search_fwdtree.c(1553): 563 words recognized (6/fr)
    INFO: ngram_search_fwdtree.c(1555): 16206 senones evaluated (169/fr)
    INFO: ngram_search_fwdtree.c(1559): 8464 channels searched (88/fr), 1086 1st, 6375 last
    INFO: ngram_search_fwdtree.c(1562): 880 words for which last channels evaluated (9/fr)
    INFO: ngram_search_fwdtree.c(1564): 228 candidate words for entering last phone (2/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.049 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.06 wall 0.057 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 5 words
    INFO: ngram_search_fwdflat.c(948): 622 words recognized (6/fr)
    INFO: ngram_search_fwdflat.c(950): 14505 senones evaluated (151/fr)
    INFO: ngram_search_fwdflat.c(952): 7629 channels searched (79/fr)
    INFO: ngram_search_fwdflat.c(954): 1101 words searched (11/fr)
    INFO: ngram_search_fwdflat.c(957): 271 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.03 CPU 0.033 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.02 wall 0.021 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .70
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 162 nodes, 295 links
    INFO: ps_lattice.c(1380): Bestpath score: -3788
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:70:94) = -256022
    INFO: ps_lattice.c(1441): Joint P(O,S) = -269561 P(S|O) = -13539
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 41.35 18.96 -26.87 35.72 -18.15 -2.71 2.54 -23.72 7.68 -11.25 -3.48 7.39 -7.41 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 40.55 19.13 -26.39 32.57 -18.31 -2.65 2.48 -22.39 7.99 -10.80 -3.10 7.16 -7.81 >
    INFO: ngram_search_fwdtree.c(1553): 387 words recognized (4/fr)
    INFO: ngram_search_fwdtree.c(1555): 13699 senones evaluated (134/fr)
    INFO: ngram_search_fwdtree.c(1559): 6805 channels searched (66/fr), 1055 1st, 5047 last
    INFO: ngram_search_fwdtree.c(1562): 756 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 189 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.046 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.05 wall 0.049 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 8 words
    INFO: ngram_search_fwdflat.c(948): 447four
    eight
    words recognized (4/fr)
    INFO: ngram_search_fwdflat.c(950): 14119 senones evaluated (138/fr)
    INFO: ngram_search_fwdflat.c(952): 8974 channels searched (87/fr)
    INFO: ngram_search_fwdflat.c(954): 1266 words searched (12/fr)
    INFO: ngram_search_fwdflat.c(957): 353 word transitions (3/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.015 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.02 wall 0.020 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .63
    INFO: ngram_search.c(1279): Eliminated 0 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 127 nodes, 151 links
    INFO: ps_lattice.c(1380): Bestpath score: -2487
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:63:100) = -185586
    INFO: ps_lattice.c(1441): Joint P(O,S) = -201402 P(S|O) = -15816
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(99): cmn_prior_update: from < 40.55 19.13 -26.39 32.57 -18.31 -2.65 2.48 -22.39 7.99 -10.80 -3.10 7.16 -7.81 >
    INFO: cmn_prior.c(116): cmn_prior_update: to < 40.62 19.30 -26.14 33.21 -18.70 -2.81 2.68 -22.94 8.20 -10.90 -3.15 6.94 -7.94 >
    INFO: cmn_prior.c(131): cmn_prior_update: from < 40.62 19.30 -26.14 33.21 -18.70 -2.81 2.68 -22.94 8.20 -10.90 -3.15 6.94 -7.94 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 38.34 16.92 -25.36 30.96 -18.14 -2.58 1.91 -21.13 7.08 -9.87 -3.41 6.60 -7.72 >
    INFO: ngram_search_fwdtree.c(1553): 359 words recognized (4/fr)
    INFO: ngram_search_fwdtree.c(1555): 11463 senones evaluated (117/fr)
    INFO: ngram_search_fwdtree.c(1559): 5454 channels searched (55/fr), 1118 1st, 3560 last
    INFO: ngram_search_fwdtree.c(1562): 736 words for which last channels evaluated (7/fr)
    INFO: ngram_search_fwdtree.c(1564): 141 candidate words for entering last phone (1/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.048 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.05 wall 0.046 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 4 words
    INFO: ngram_search_fwdflat.c(948): 448 words recognized (5/fr)
    INFO: ngram_search_fwdflat.c(950): 8433 senones evaluated (86/fr)
    INFO: ngram_search_fwdflat.c(952): 3914 channels searched (39/fr)
    INFO: ngram_search_fwdflat.c(954): 990 words searched (10/fr)
    INFO: ngram_search_fwdflat.c(957): 196 word transitions (2/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.02 CPU 0.016 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.01 wall 0.010 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .41
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 85 nodes, 64 links
    INFO: ps_lattice.c(1380): Bestpath score: -1706
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:41:96) = -167245
    INFO: ps_lattice.c(1441): Joint P(O,S) = -200318 P(S|O) = -33073
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: cmn_prior.c(131): cmn_prior_update: from < 38.34 16.92 -25.36 30.96 -18.14 -2.58 1.91 -21.13 7.08 -9.87 -3.41 6.60 -7.72 >
    INFO: cmn_prior.c(149): cmn_prior_update: to < 39.15 17.59 -27.05 29.90 -18.46 -2.14 2.83 -21.15 8.23 -9.13 -3.93 6.50 -7.13 >
    INFO: ngram_search_fwdtree.c(1553): 556 words recognized (7/fr)
    INFO: ngram_search_fwdtree.c(1555): 15221 senones evaluated (179/fr)
    INFO: ngram_search_fwdtree.c(1559): 8953 channels searched (105/fr), 972 1st, 7144 last
    INFO: ngram_search_fwdtree.c(1562): 844 words for which last channels evaluated (9/fr)
    INFO: ngram_search_fwdtree.c(1564): 250 candidate words for entering last phone (2/fr)
    INFO: ngram_search_fwdtree.c(1567): fwdtree 0.05 CPU 0.055 xRT
    INFO: ngram_search_fwdtree.c(1570): fwdtree 0.05 wall 0.059 xRT
    INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 8 words
    INFO: ngram_search_fwdflat.c(948): 523 words recognized (6/fr)
    INFO: ngram_search_fwdflat.c(950): 15344 senones evaluated (181/fr)
    INFO: ngram_search_fwdflat.c(952): 10863 channels seanine
    rched (127/fr)
    INFO: ngram_search_fwdflat.c(954): 1177 words searched (13/fr)
    INFO: ngram_search_fwdflat.c(957): 346 word transitions (4/fr)
    INFO: ngram_search_fwdflat.c(960): fwdflat 0.03 CPU 0.037 xRT
    INFO: ngram_search_fwdflat.c(963): fwdflat 0.03 wall 0.035 xRT
    INFO: ngram_search.c(1253): lattice start node .0 end node .60
    INFO: ngram_search.c(1279): Eliminated 1 nodes before end node
    INFO: ngram_search.c(1384): Lattice has 134 nodes, 336 links
    INFO: ps_lattice.c(1380): Bestpath score: -2644
    INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(:60:83) = -151454
    INFO: ps_lattice.c(1441): Joint P(O,S) = -210101 P(S|O) = -58647
    INFO: ngram_search.c(875): bestpath 0.00 CPU 0.000 xRT
    INFO: ngram_search.c(878): bestpath 0.00 wall 0.000 xRT
    INFO: ngram_search_fwdtree.c(432): TOTAL fwdtree 1.70 CPU 0.058 xRT
    INFO: ngram_search_fwdtree.c(435): TOTAL fwdtree 1.72 wall 0.059 xRT
    INFO: ngram_search_fwdflat.c(176): TOTAL fwdflat 1.08 CPU 0.037 xRT
    INFO: ngram_search_fwdflat.c(179): TOTAL fwdflat 1.05 wall 0.036 xRT
    INFO: ngram_search.c(303): TOTAL bestpath 0.02 CPU 0.001 xRT
    INFO: ngram_search.c(306): TOTAL bestpath 0.02 wall 0.001 xRT

    The output at this point:

    eight two
    eight two two one
    nine three one two one two one three
    three
    eight
    three
    eight
    three
    four
    eight
    eight
    eight
    three
    eight
    three
    four
    eight
    nine

    When the output should actually be:

    eight two
    eight two two one
    nine three one two one two one three
    three
    six
    three
    eight
    three
    four
    eight
    eight
    six
    three
    eight
    three
    four
    eight
    nine

    I have no idea, how or why, it's confusing the "eight" with the "six", I should mention that as you could see on the dictionary I tried to add more references into the six, just in case that pronounciation wasn't correct, including similar sounds, however, I'm still getting the same result.

    I understand that the basic for a good accuracy are basically 3 parameters:

    -hmm model/en-us/en-us
    -lm model/en-us/en-us.lm.bin
    -dict model/en-us/micro.dict

    As my knowledge on this field is quite limited, I was only able to remove a lot of words from the original dictionary, creating a smaller one. Doing this I was able to recognize most of the numbers, all of them actually, with exeption of the six. I also downloaded and tested different hmm (models?) and lm files (.bin), but didn't change too much...

    Any help would be really appreciated!
    Kind regards,

     

  • Chris C. Russo

    Chris C. Russo - 2016-03-21

    03:13 AM:

    The feat.params file that I'm using has the -nfilt value originally defined in 25. Now, I realized, that if I decrease the value, to 5 for example...

    Almost every single number becomes "eight":

    one two
    one one one one one one
    eight
    eight
    eight
    eight
    eight
    eight
    eight
    eight
    eight
    eight

    I imagine that this is somehow related with the problem, so I'm trying to increase the value and see if I can fix this. The manual states that nfilt is "The number of filter banks", I'm trying to comprehend what exactly does a filter bank...

    Help would be more than welcome! I hope this post helps someone else as well!

     

    • Nickolay V. Shmyrev

      Nickolay V. Shmyrev - 2016-03-21

      Hello Chris

      1) Don't touch paramters you do not understand. You should never change nfilt
      2) Don't modify the dictionary, modify language model instead as described in our tutorial
      3) If you need help on the accuracy provide data files (recordings) to reproduce your problems.

       

  • Chris C. Russo

    Chris C. Russo - 2016-03-21

    Hello Nickolay, Thanks a lot for the quick feedback!

    Regarding the point number (3) described in your approach, is there any way to prepare the language model without many samples?

     

    • Nickolay V. Shmyrev

      Nickolay V. Shmyrev - 2016-03-21

      Hello Chris

      In point 3) I never wrote you need many samples. Actually langauge model training does not require you to provide samples, you can just list the words you expect to see in a grammar. For more details please read the tutorial

      http://cmusphinx.sourceforge.net/wiki/tutoriallm

       

  • Chris C. Russo

    Chris C. Russo - 2016-03-25

    Hello Nickolay, thanks a lot for your help so far.

    I've been working on this like a maniac, but I'm pretty sure that I'm missing something because I'm not being able to recognize a few digits.

    I tried the grammar file as you suggested:

    JSGF V1.0;
    grammar numbers;
    public <number> = zero | one | two | three | four | five | six | seven | eight | nine ;</number>

    I tried also reducing the dictionary as I mentioned, as the only thing I need are numbers:

    zero Z IH R OW
    zero(2) Z IY R OW
    one W AH N
    one(2) HH W AH N
    two T UW
    three TH R IY
    four F AO R
    five F AY V
    six S F IH K S
    seven S EH V AH N
    eight EY T
    nine N AY N

    And finally I also downloaded an uncompressed (non-bin) .lm file and modified it to fit:

    This is an ARPA-format language model file, generated by CMU Sphinx
    \data\
    ngram 1=12

    \1-grams:
    -1.4016 -0.4051
    -99.0000 1.4124
    -3.8842 zero -0.4376
    -2.4976 one -0.4239
    -2.7508 two -0.2802
    -2.9571 three -0.3222
    -3.0738 four -0.3780
    -2.9962 five -0.4140
    -3.1903 six -0.3997
    -3.2768 seven -0.3833
    -3.3046 eight -0.3988
    -3.4077 nine -0.3758

    \end\

    Before modifying anything at all, I tried the different components directly out of the box, with very similar accuracy as well.

    **But the problem persists! **

    4 different .wav files pronouncing clearly "475125" (...four... ...seven... ...five... ...one... ...two... ...five), for some reason, it heards "175125" (the first 4 is proccessed as 1).

    A second group of audio files, where the number dictated is "542757", is recognized as "512757", once again, everything looks normal but the 4 is proccessed as 1.

    I really have no idea where else I could check..

    Models:

    I'm using "cmusphinx-en-us-5.2".

    I'm seriously without a clue, walking on air at this point...

    Sox conversion proccess:

      $sox_command        = "$sox_bin $wav_file ";
  $sox_command        .= "-b 16 ";
  $sox_command        .= "$wav_conversion ";
  $sox_command        .= "channels 1 ";
  $sox_command        .= "rate 16000 ";
  $sox_command        .= "dither -a ";
  $sox_command        .= "-f shibata ";
  $sox_command        .= "trim 00:13 00:30";

  //lanzamos sox
  system("$sox_command");

    Sphinx exec:

      $sphinx_command     = "$sphinxc_bin -infile $wav_conversion ";
  $sphinx_command     .= "-hmm $sphinxc_hmm ";
  $sphinx_command     .= "-lm $sphinxc_lm ";
  $sphinx_command     .= "-dict $sphinxc_dic ";
  $sphinx_command     .= "-jsgf $sphinxc_gram ";
  $sphinx_command     .= "-logfn $sphinxc_records ";
  $sphinx_command     .= "-lw 1 ";
  $sphinx_command     .= "-remove_dc yes ";
  $sphinx_command     .= "-remove_noise yes ";
  $sphinx_command     .= "-vad_threshold 2.4 ";
  $sphinx_command     .= "-vad_prespeech 19 ";
  $sphinx_command     .= "-vad_postspeech 37 ";
  $sphinx_command     .= "-silprob 3.5 ";
  $sphinx_command     .= "> $sphinxc_feedback";

  //lanzamos sphinx
  system("$sphinx_command");

    I'm adding some sample files... hope someone has some ideas!

    1,000 Thanks!
    Chris C.

     

    Last edit: Chris C. Russo 2016-03-25

  • Chris C. Russo

    Chris C. Russo - 2016-03-25

    Ok, I found a really nice one:

    If I run from mic:

    sphinxc.exe -inmic yes -hmm cmusphinx-en-us-5.2 -lm en-us.lm -dict micro.dict -jsgf grammar.txt -logfn record.txt -lw 1 -remove_dc yes -remove_noise yes -vad_threshold 2.4 -vad_prespeech 19 -vad_postspeech 37 -silprob 3.5

    Place mic near speakers and play wav on any audio player. Works flawlessly. WER 0%

    If I run from file:

    sphinxc.exe -infile conversion_2033910124_1458812700.wav -hmm cmusphinx-en-us-5.2 -lm en-us.lm -dict micro.dict -jsgf grammar.txt -logfn record.txt -lw 1 -remove_dc yes -remove_noise yes -vad_threshold 2.4 -vad_prespeech 19 -vad_postspeech 37 -silprob 3.5

    4 becomes 1.

    Same command, same file, from mic works (playback from VLC / Winamp), using sphinx, fails.

     

Log in to post a comment.