Different accuracy for continuous and batch
ilichev-andrey opened this issue · 5 comments
ilichev-andrey commented
Good afternoon.
I ran a test for pocketsphinx_batch and pocketsphinx_continuous and got different results.
Input parameters are identical!
For a set of 127 files, the results were:
pocketsphinx_continuous:
TOTAL Words: 990 Correct: 477 Errors: 641
TOTAL Percent correct = 48.18% Error = 64.75% Accuracy = 35.25%
TOTAL Insertions: 128 Deletions: 32 Substitutions: 481
pocketsphinx_batch:
TOTAL Words: 990 Correct: 821 Errors: 194
TOTAL Percent correct = 82.93% Error = 19.60% Accuracy = 80.40%
TOTAL Insertions: 25 Deletions: 20 Substitutions: 149
For demonstration this file is taken.
wav file - 4 second (samprate 8000)
pocketsphinx_continuous:
Words: 9 Correct: 2 Errors: 7 Percent correct = 22.22% Error = 77.78% Accuracy = 22.22%
Insertions: 0 Deletions: 1 Substitutions: 6
pocketsphinx_batch:
Words: 9 Correct: 8 Errors: 2 Percent correct = 88.89% Error = 22.22% Accuracy = 77.78%
Insertions: 1 Deletions: 0 Substitutions: 1
pocketsphinx_continuous:
INFO: pocketsphinx.c(152): Parsed model-specific feature parameters from /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/feat.params
Current configuration:
[NAME] [DEFLT] [VALUE]
-agc none none
-agcthresh 2.0 2.000000e+00
-allphone
-allphone_ci no no
-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
-ceplen 13 13
-cmn live current
-cmninit 40,3,-1 11.64,0.15,-0.04,0.17,-0.40,-0.03,-0.50,-0.13,-0.33,-0.11,-0.19,-0.10,-0.24
-compallsen no no
-debug 0
-dict /root/cmu_sphinx/demo_custom/var8/6_combine_dict/mixed.data.ru.dic
-dictcase no no
-dither no yes
-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 /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000
-input_endian little little
-jsgf
-keyphrase
-kws
-kws_delay 10 10
-kws_plp 1e-1 1.000000e-01
-kws_threshold 1 1.000000e+00
-latsize 5000 5000
-lda
-ldadim 0 0
-lifter 0 0
-lm /root/cmu_sphinx/demo_custom/var8/5_combine_arpa_srilm/mixed_data.kndiscount.interpolate_ru.lm
-lmctl
-lmname
-logbase 1.0001 1.000100e+00
-logfn 00004.wav.log
-logspec no no
-lowerf 133.33334 1.300000e+02
-lpbeam 1e-40 1.000000e-40
-lponlybeam 7e-29 7.000000e-29
-lw 6.5 6.500000e+00
-maxhmmpf 30000 30000
-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 31
-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-10 1.000000e-10
-pl_pip 1.0 1.000000e+00
-pl_weight 3.0 3.000000e+00
-pl_window 5 5
-rawlogdir
-remove_dc no no
-remove_noise yes no
-remove_silence yes yes
-round_filters yes yes
-samprate 16000 8.000000e+03
-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 3.700000e+03
-uw 1.0 1.000000e+00
-vad_postspeech 50 50
-vad_prespeech 20 20
-vad_startspeech 10 10
-vad_threshold 2.0 2.000000e+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
pocketsphinx_batch:
INFO: pocketsphinx.c(152): Parsed model-specific feature parameters from /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/feat.params
Current configuration:
[NAME] [DEFLT] [VALUE]
-agc none none
-agcthresh 2.0 2.000000e+00
-allphone
-allphone_ci no no
-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
-ceplen 13 13
-cmn live current
-cmninit 40,3,-1 11.64,0.15,-0.04,0.17,-0.40,-0.03,-0.50,-0.13,-0.33,-0.11,-0.19,-0.10,-0.24
-compallsen no no
-debug 0
-dict /root/cmu_sphinx/demo_custom/var8/6_combine_dict/mixed.data.ru.dic
-dictcase no no
-dither no yes
-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 /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000
-input_endian little little
-jsgf
-keyphrase
-kws
-kws_delay 10 10
-kws_plp 1e-1 1.000000e-01
-kws_threshold 1 1.000000e+00
-latsize 5000 5000
-lda
-ldadim 0 0
-lifter 0 0
-lm /root/cmu_sphinx/demo_custom/var8/5_combine_arpa_srilm/mixed_data.kndiscount.interpolate_ru.lm
-lmctl
-lmname
-logbase 1.0001 1.000100e+00
-logfn phrases.log
-logspec no no
-lowerf 133.33334 1.300000e+02
-lpbeam 1e-40 1.000000e-40
-lponlybeam 7e-29 7.000000e-29
-lw 6.5 6.500000e+00
-maxhmmpf 30000 30000
-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 31
-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-10 1.000000e-10
-pl_pip 1.0 1.000000e+00
-pl_weight 3.0 3.000000e+00
-pl_window 5 5
-rawlogdir
-remove_dc no no
-remove_noise yes no
-remove_silence yes yes
-round_filters yes yes
-samprate 16000 8.000000e+03
-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 3.700000e+03
-uw 1.0 1.000000e+00
-vad_postspeech 50 50
-vad_prespeech 20 20
-vad_startspeech 10 10
-vad_threshold 2.0 2.000000e+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
pocketsphinx_continuous:
NFO: fe_interface.c(324): Using -1 as the seed.
INFO: feat.c(715): Initializing feature stream to type: '1s_c_d_dd', ceplen=13, CMN='batch', VARNORM='no', AGC='none'
INFO: acmod.c(152): Reading linear feature transformation from /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/feature_transform
INFO: mdef.c(518): Reading model definition: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/mdef
INFO: bin_mdef.c(181): Allocating 145321 * 8 bytes (1135 KiB) for CD tree
INFO: tmat.c(149): Reading HMM transition probability matrices: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/transition_matrices
INFO: acmod.c(113): Attempting to use PTM computation module
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: ptm_mgau.c(804): Number of codebooks exceeds 256: 4159
INFO: acmod.c(115): Attempting to use semi-continuous computation module
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: acmod.c(117): Falling back to general multi-stream GMM computation
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: ms_senone.c(149): Reading senone mixture weights: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/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 4159 senones: 1 features x 63 codewords
INFO: ms_senone.c(320): Mapping senones to individual codebooks
INFO: ms_mgau.c(144): 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 523274 * 32 bytes (16352 KiB) for word entries
INFO: dict.c(333): Reading main dictionary: /root/cmu_sphinx/demo_custom/var8/6_combine_dict/mixed.data.ru.dic
INFO: dict.c(213): Dictionary size 519175, allocated 8831 KiB for strings, 8547 KiB for phones
INFO: dict.c(336): 519175 words read
INFO: dict.c(358): Reading filler dictionary: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/noisedict
INFO: dict.c(213): Dictionary size 519178, allocated 0 KiB for strings, 0 KiB for phones
INFO: dict.c(361): 3 words read
INFO: dict2pid.c(396): Building PID tables for dictionary
INFO: dict2pid.c(406): Allocating 53^3 * 2 bytes (290 KiB) for word-initial triphones
INFO: dict2pid.c(132): Allocated 67840 bytes (66 KiB) for word-final triphones
INFO: dict2pid.c(196): Allocated 67840 bytes (66 KiB) for single-phone word triphones
INFO: ngram_model_trie.c(354): Trying to read LM in trie binary format
INFO: ngram_model_trie.c(365): Header doesn't match
INFO: ngram_model_trie.c(177): Trying to read LM in arpa format
INFO: ngram_model_trie.c(193): LM of order 3
INFO: ngram_model_trie.c(195): #1-grams: 509309
INFO: ngram_model_trie.c(195): #2-grams: 7261937
INFO: ngram_model_trie.c(195): #3-grams: 4988627
INFO: lm_trie.c(474): Training quantizer
INFO: lm_trie.c(482): Building LM trie
INFO: ngram_search_fwdtree.c(74): Initializing search tree
INFO: ngram_search_fwdtree.c(101): 1247 unique initial diphones
INFO: ngram_search_fwdtree.c(186): Creating search channels
INFO: ngram_search_fwdtree.c(323): Max nonroot chan increased to 915818
INFO: ngram_search_fwdtree.c(333): Created 1247 root, 915690 non-root channels, 85 single-phone words
INFO: ngram_search_fwdflat.c(157): fwdflat: min_ef_width = 4, max_sf_win = 25
INFO: ngram_search.c(467): Resized score stack to 200000 entries
INFO: ngram_search.c(459): Resized backpointer table to 10000 entries
INFO: ngram_search.c(467): Resized score stack to 400000 entries
INFO: ngram_search.c(459): Resized backpointer table to 20000 entries
INFO: ngram_search.c(467): Resized score stack to 800000 entries
INFO: ngram_search.c(459): Resized backpointer table to 40000 entries
INFO: ngram_search.c(467): Resized score stack to 1600000 entries
INFO: cmn_live.c(120): Update from < 11.64 0.15 -0.04 0.17 -0.40 -0.03 -0.50 -0.13 -0.33 -0.11 -0.19 -0.10 -0.24 >
INFO: cmn_live.c(138): Update to < 15.05 -0.54 0.14 -0.17 -0.56 -0.34 -0.30 -0.41 -0.23 -0.27 -0.19 -0.23 -0.23 >
INFO: ngram_search_fwdtree.c(1550): 28833 words recognized (68/fr)
INFO: ngram_search_fwdtree.c(1552): 994179 senones evaluated (2356/fr)
INFO: ngram_search_fwdtree.c(1556): 5247759 channels searched (12435/fr), 228163 1st, 968395 last
INFO: ngram_search_fwdtree.c(1559): 48812 words for which last channels evaluated (115/fr)
INFO: ngram_search_fwdtree.c(1561): 263511 candidate words for entering last phone (624/fr)
INFO: ngram_search_fwdtree.c(1564): fwdtree 4.70 CPU 1.113 xRT
INFO: ngram_search_fwdtree.c(1567): fwdtree 4.70 wall 1.113 xRT
INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 1984 words
INFO: ngram_search_fwdflat.c(948): 16665 words recognized (39/fr)
INFO: ngram_search_fwdflat.c(950): 569855 senones evaluated (1350/fr)
INFO: ngram_search_fwdflat.c(952): 2090701 channels searched (4954/fr)
INFO: ngram_search_fwdflat.c(954): 139195 words searched (329/fr)
INFO: ngram_search_fwdflat.c(957): 110745 word transitions (262/fr)
INFO: ngram_search_fwdflat.c(960): fwdflat 2.11 CPU 0.500 xRT
INFO: ngram_search_fwdflat.c(963): fwdflat 2.11 wall 0.499 xRT
INFO: ngram_search.c(1250): lattice start node <s>.0 end node </s>.405
INFO: ngram_search.c(1276): Eliminated 0 nodes before end node
INFO: ngram_search.c(1381): Lattice has 1928 nodes, 21510 links
INFO: ps_lattice.c(1380): Bestpath score: -28323
INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(</s>:405:420) = -1773292
INFO: ps_lattice.c(1441): Joint P(O,S) = -1923487 P(S|O) = -150195
INFO: ngram_search.c(872): bestpath 0.10 CPU 0.025 xRT
INFO: ngram_search.c(875): bestpath 0.11 wall 0.025 xRT
INFO: ngram_search.c(1027): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(1030): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search.c(1027): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(1030): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search.c(872): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(875): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search.c(872): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(875): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search.c(872): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(875): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search.c(872): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(875): bestpath 0.00 wall 0.000 xRT
INFO: ngram_search_fwdtree.c(429): TOTAL fwdtree 4.70 CPU 1.115 xRT
INFO: ngram_search_fwdtree.c(432): TOTAL fwdtree 4.70 wall 1.116 xRT
INFO: ngram_search_fwdflat.c(176): TOTAL fwdflat 2.11 CPU 0.501 xRT
INFO: ngram_search_fwdflat.c(179): TOTAL fwdflat 2.11 wall 0.500 xRT
INFO: ngram_search.c(303): TOTAL bestpath 0.10 CPU 0.025 xRT
INFO: ngram_search.c(306): TOTAL bestpath 0.11 wall 0.025 xRT
pocketsphinx_batch:
INFO: fe_interface.c(324): Using -1 as the seed.
INFO: feat.c(715): Initializing feature stream to type: '1s_c_d_dd', ceplen=13, CMN='batch', VARNORM='no', AGC='none'
INFO: acmod.c(152): Reading linear feature transformation from /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/feature_transform
INFO: mdef.c(518): Reading model definition: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/mdef
INFO: bin_mdef.c(181): Allocating 145321 * 8 bytes (1135 KiB) for CD tree
INFO: tmat.c(149): Reading HMM transition probability matrices: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/transition_matrices
INFO: acmod.c(113): Attempting to use PTM computation module
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: ptm_mgau.c(804): Number of codebooks exceeds 256: 4159
INFO: acmod.c(115): Attempting to use semi-continuous computation module
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: acmod.c(117): Falling back to general multi-stream GMM computation
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/means
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(127): Reading mixture gaussian parameter: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/variances
INFO: ms_gauden.c(242): 4159 codebook, 1 feature, size:
INFO: ms_gauden.c(244): 63x29
INFO: ms_gauden.c(304): 182668 variance values floored
INFO: ms_senone.c(149): Reading senone mixture weights: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/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 4159 senones: 1 features x 63 codewords
INFO: ms_senone.c(320): Mapping senones to individual codebooks
INFO: ms_mgau.c(144): 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 523274 * 32 bytes (16352 KiB) for word entries
INFO: dict.c(333): Reading main dictionary: /root/cmu_sphinx/demo_custom/var8/6_combine_dict/mixed.data.ru.dic
INFO: dict.c(213): Dictionary size 519175, allocated 8831 KiB for strings, 8547 KiB for phones
INFO: dict.c(336): 519175 words read
INFO: dict.c(358): Reading filler dictionary: /root/cmu_sphinx/demo_custom/hmm/zero_ru_cont_8k_v3/zero_ru.cd_cont_4000/noisedict
INFO: dict.c(213): Dictionary size 519178, allocated 0 KiB for strings, 0 KiB for phones
INFO: dict.c(361): 3 words read
INFO: dict2pid.c(396): Building PID tables for dictionary
INFO: dict2pid.c(406): Allocating 53^3 * 2 bytes (290 KiB) for word-initial triphones
INFO: dict2pid.c(132): Allocated 67840 bytes (66 KiB) for word-final triphones
INFO: dict2pid.c(196): Allocated 67840 bytes (66 KiB) for single-phone word triphones
INFO: ngram_model_trie.c(354): Trying to read LM in trie binary format
INFO: ngram_model_trie.c(365): Header doesn't match
INFO: ngram_model_trie.c(177): Trying to read LM in arpa format
INFO: ngram_model_trie.c(193): LM of order 3
INFO: ngram_model_trie.c(195): #1-grams: 509309
INFO: ngram_model_trie.c(195): #2-grams: 7261937
INFO: ngram_model_trie.c(195): #3-grams: 4988627
INFO: lm_trie.c(474): Training quantizer
INFO: lm_trie.c(482): Building LM trie
INFO: ngram_search_fwdtree.c(74): Initializing search tree
INFO: ngram_search_fwdtree.c(101): 1247 unique initial diphones
INFO: ngram_search_fwdtree.c(186): Creating search channels
INFO: ngram_search_fwdtree.c(323): Max nonroot chan increased to 915818
INFO: ngram_search_fwdtree.c(333): Created 1247 root, 915690 non-root channels, 85 single-phone words
INFO: ngram_search_fwdflat.c(157): fwdflat: min_ef_width = 4, max_sf_win = 25
INFO: batch.c(729): Decoding '00004'
INFO: cmn.c(133): CMN: 15.09 -0.54 0.14 -0.17 -0.56 -0.34 -0.30 -0.41 -0.23 -0.27 -0.19 -0.23 -0.23
INFO: ngram_search_fwdtree.c(1550): 14156 words recognized (34/fr)
INFO: ngram_search_fwdtree.c(1552): 1026440 senones evaluated (2438/fr)
INFO: ngram_search_fwdtree.c(1556): 5226094 channels searched (12413/fr), 322875 1st, 552390 last
INFO: ngram_search_fwdtree.c(1559): 27663 words for which last channels evaluated (65/fr)
INFO: ngram_search_fwdtree.c(1561): 247689 candidate words for entering last phone (588/fr)
INFO: ngram_search_fwdtree.c(1564): fwdtree 4.96 CPU 1.179 xRT
INFO: ngram_search_fwdtree.c(1567): fwdtree 4.96 wall 1.179 xRT
INFO: ngram_search_fwdflat.c(302): Utterance vocabulary contains 1077 words
INFO: ngram_search_fwdflat.c(948): 6467 words recognized (15/fr)
INFO: ngram_search_fwdflat.c(950): 366876 senones evaluated (871/fr)
INFO: ngram_search_fwdflat.c(952): 900909 channels searched (2139/fr)
INFO: ngram_search_fwdflat.c(954): 64443 words searched (153/fr)
INFO: ngram_search_fwdflat.c(957): 60051 word transitions (142/fr)
INFO: ngram_search_fwdflat.c(960): fwdflat 1.38 CPU 0.328 xRT
INFO: ngram_search_fwdflat.c(963): fwdflat 1.38 wall 0.328 xRT
INFO: ngram_search.c(1250): lattice start node <s>.0 end node </s>.409
INFO: ngram_search.c(1276): Eliminated 0 nodes before end node
INFO: ngram_search.c(1381): Lattice has 1045 nodes, 1411 links
INFO: ps_lattice.c(1380): Bestpath score: -20287
INFO: ps_lattice.c(1384): Normalizer P(O) = alpha(</s>:409:419) = -1376598
INFO: ps_lattice.c(1441): Joint P(O,S) = -1426223 P(S|O) = -49625
INFO: ngram_search.c(872): bestpath 0.01 CPU 0.002 xRT
INFO: ngram_search.c(875): bestpath 0.01 wall 0.001 xRT
INFO: batch.c(761): 00004: 4.20 seconds speech, 6.35 seconds CPU, 6.35 seconds wall
INFO: batch.c(763): 00004: 1.51 xRT (CPU), 1.51 xRT (elapsed)
nshmyrev commented
ilichev-andrey commented
For my dataset, the following solution helped:
From: #63 (comment)
- your audio is pretty short, you need batch processing for higher accuracy. You need to process whole audio at once with ps_process_raw with last argument full_utt = TRUE. See discussion too.
My dataset:
127 wav files (from 1 to 5 seccond)
KiAlexander commented
so how to improve the performance by ps_continuous as by ps_batch? Even I set the same cmninit as ps_batch ,the result is still terrible.
nshmyrev commented
@KiAlexander avoid updating old unrealted issues. Share the data to reproduce your problems to get help.
mainvolume commented