Sphinx武林秘籍(下)

2016/06/30 15:20
阅读数 65

 

Sphinx武林秘籍()

 

――使用训练好的语言模型与声学模型

 

一、第一次使用 
#cp -rf my_db.cd_cont_1000 /usr/local/bin
#cd .. 
#cd etc 
#cp my_db.dic my_db.lm.DMP /usr/local/bin/ 
#cd /usr/local/bin 

 

# ./pocketsphinx_continuous -hmm my_db.cd_cont_1000 -lm my_db.lm.DMP -dict my_db.dic

 

 

 

INFO: cmd_ln.c(512): Parsing command line:

 

./pocketsphinx_continuous \

 

       -hmm my_db.cd_cont_1000 \

 

       -lm my_db.lm.DMP \

 

       -dict my_db.dic

 

 

 

Current configuration:

 

[NAME]        [DEFLT]        [VALUE]

 

-adcdev                       

 

-agc        none              none

 

-agcthresh       2.0          2.000000e+00

 

-alpha            0.97        9.700000e-01

 

-argfile                 

 

-ascale            20.0        2.000000e+01

 

-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                      my_db.dic

 

-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                           my_db.cd_cont_1000

 

-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                       my_db.lm.DMP

 

-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                   

 

-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                    

 

-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(512): Parsing command line:

 

\

 

       -alpha 0.97 \

 

       -dither yes \

 

       -doublebw no \

 

       -nfilt 40 \

 

       -ncep 13 \

 

       -lowerf 133.33334 \

 

       -upperf 6855.4976 \

 

       -nfft 512 \

 

       -wlen 0.0256 \

 

       -transform legacy \

 

       -feat 1s_c_d_dd \

 

       -agc none \

 

       -cmn current \

 

       -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          8.0

 

-dither            no           yes

 

-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      1.333333e+02

 

-ncep             13           13

 

-nfft        512         512

 

-nfilt              40           40

 

-remove_dc    no           no

 

-round_filters  yes          yes

 

-samprate       16000            1.600000e+04

 

-seed              -1           -1

 

-smoothspec    no           no

 

-svspec                        

 

-transform      legacy            legacy

 

-unit_area       yes          yes

 

-upperf           6855.4976      6.855498e+03

 

-varnorm no           no

 

-verbose  no           no

 

-warp_params              

 

-warp_type     inverse_linear inverse_linear

 

-wlen             0.025625 2.560000e-02

 

 

 

INFO: acmod.c(238): Parsed model-specific feature parameters from my_db.cd_cont_1000/feat.params

 

INFO: fe_interface.c(288): You are using the internal mechanism to generate the seed.

 

INFO: feat.c(848): 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: mdef.c(520): Reading model definition: my_db.cd_cont_1000/mdef

 

INFO: bin_mdef.c(173): Allocating 304 * 8 bytes (2 KiB) for CD tree

 

INFO: tmat.c(205): Reading HMM transition probability matrices: my_db.cd_cont_1000/transition_matrices

 

INFO: acmod.c(117): Attempting to use SCHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 30781 variance values floored

 

INFO: acmod.c(119): Attempting to use PTHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 30781 variance values floored

 

INFO: ptm_mgau.c(671): Reading mixture weights file 'my_db.cd_cont_1000/mixture_weights'

 

INFO: ptm_mgau.c(765): Read 105 x 1 x 8 mixture weights

 

INFO: ptm_mgau.c(831): Maximum top-N: 4

 

INFO: dict.c(294): Allocating 4112 * 20 bytes (80 KiB) for word entries

 

INFO: dict.c(306): Reading main dictionary: my_db.dic

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(309): 13 words read

 

INFO: dict.c(314): Reading filler dictionary: my_db.cd_cont_1000/noisedict

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(317): 3 words read

 

INFO: dict2pid.c(396): Building PID tables for dictionary

 

INFO: dict2pid.c(405): Allocating 16^3 * 2 bytes (8 KiB) for word-initial triphones

 

INFO: dict2pid.c(131): Allocated 3136 bytes (3 KiB) for word-final triphones

 

INFO: dict2pid.c(195): Allocated 3136 bytes (3 KiB) for single-phone word triphones

 

ERROR: "ngram_model_arpa.c", line 76: No \data\ mark in LM file

 

INFO: ngram_model_dmp.c(141): Will use memory-mapped I/O for LM file

 

INFO: ngram_model_dmp.c(195): ngrams 1=8, 2=10, 3=13

 

INFO: ngram_model_dmp.c(241):        8 = LM.unigrams(+trailer) read

 

INFO: ngram_model_dmp.c(289):       10 = LM.bigrams(+trailer) read

 

INFO: ngram_model_dmp.c(314):       13 = LM.trigrams read

 

INFO: ngram_model_dmp.c(338):        4 = LM.prob2 entries read

 

INFO: ngram_model_dmp.c(357):        5 = LM.bo_wt2 entries read

 

INFO: ngram_model_dmp.c(377):        3 = LM.prob3 entries read

 

INFO: ngram_model_dmp.c(405):        1 = LM.tseg_base entries read

 

INFO: ngram_model_dmp.c(461):        8 = ascii word strings read

 

INFO: ngram_search_fwdtree.c(99): 8 unique initial diphones

 

INFO: ngram_search_fwdtree.c(147): 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(186): Creating search tree

 

INFO: ngram_search_fwdtree.c(191): before: 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(324): after: max nonroot chan increased to 138

 

INFO: ngram_search_fwdtree.c(333): after: 5 root, 10 non-root channels, 3 single-phone words

 

INFO: ngram_search_fwdflat.c(153): fwdflat: min_ef_width = 4, max_sf_win = 25

 

Warning: Could not find Mic element

 

INFO: continuous.c(261): ./pocketsphinx_continuous COMPILED ON: Feb 21 2011, AT: 22:31:47

 

 

 

READY....

 

 

 

错误: ERROR: "ngram_model_arpa.c", line 76: No \data\ mark in LM file 可忽略跳过

 

警告: Warning: Could not find Mic element 提示找不到麦克。。。

 

修正执行命令:./pocketsphinx_continuous -adcdev hw:AudioPCI -hmm my_db.cd_cont_1000 -lm my_db.lm.DMP -dict my_db.dic

 

二、第二次

 

#./pocketsphinx_continuous -adcdev hw:AudioPCI -hmm my_db.cd_cont_1000 -lm my_db.lm.DMP -dict my_db.dic

 

INFO: cmd_ln.c(512): Parsing command line:

 

./pocketsphinx_continuous \

 

       -hmm my_db.cd_cont_1000 \

 

       -lm my_db.lm.DMP \

 

       -dict my_db.dic

 

 

 

Current configuration:

 

[NAME]        [DEFLT]        [VALUE]

 

-adcdev                       

 

-agc        none              none

 

-agcthresh       2.0          2.000000e+00

 

-alpha            0.97        9.700000e-01

 

-argfile                 

 

-ascale            20.0        2.000000e+01

 

-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                      my_db.dic

 

-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                           my_db.cd_cont_1000

 

-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                       my_db.lm.DMP

 

-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                   

 

-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                    

 

-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(512): Parsing command line:

 

\

 

       -alpha 0.97 \

 

       -dither yes \

 

       -doublebw no \

 

       -nfilt 40 \

 

       -ncep 13 \

 

       -lowerf 133.33334 \

 

       -upperf 6855.4976 \

 

       -nfft 512 \

 

       -wlen 0.0256 \

 

       -transform legacy \

 

       -feat 1s_c_d_dd \

 

       -agc none \

 

       -cmn current \

 

       -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          8.0

 

-dither            no           yes

 

-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      1.333333e+02

 

-ncep             13           13

 

-nfft        512         512

 

-nfilt              40           40

 

-remove_dc    no           no

 

-round_filters  yes          yes

 

-samprate       16000            1.600000e+04

 

-seed              -1           -1

 

-smoothspec    no           no

 

-svspec                        

 

-transform      legacy            legacy

 

-unit_area       yes          yes

 

-upperf           6855.4976      6.855498e+03

 

-varnorm no           no

 

-verbose  no           no

 

-warp_params              

 

-warp_type     inverse_linear inverse_linear

 

-wlen             0.025625 2.560000e-02

 

 

 

INFO: acmod.c(238): Parsed model-specific feature parameters from my_db.cd_cont_1000/feat.params

 

INFO: fe_interface.c(288): You are using the internal mechanism to generate the seed.

 

INFO: feat.c(848): 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: mdef.c(520): Reading model definition: my_db.cd_cont_1000/mdef

 

INFO: bin_mdef.c(173): Allocating 304 * 8 bytes (2 KiB) for CD tree

 

INFO: tmat.c(205): Reading HMM transition probability matrices: my_db.cd_cont_1000/transition_matrices

 

INFO: acmod.c(117): Attempting to use SCHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 30781 variance values floored

 

INFO: acmod.c(119): Attempting to use PTHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 30781 variance values floored

 

INFO: ptm_mgau.c(671): Reading mixture weights file 'my_db.cd_cont_1000/mixture_weights'

 

INFO: ptm_mgau.c(765): Read 105 x 1 x 8 mixture weights

 

INFO: ptm_mgau.c(831): Maximum top-N: 4

 

INFO: dict.c(294): Allocating 4112 * 20 bytes (80 KiB) for word entries

 

INFO: dict.c(306): Reading main dictionary: my_db.dic

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(309): 13 words read

 

INFO: dict.c(314): Reading filler dictionary: my_db.cd_cont_1000/noisedict

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(317): 3 words read

 

INFO: dict2pid.c(396): Building PID tables for dictionary

 

INFO: dict2pid.c(405): Allocating 16^3 * 2 bytes (8 KiB) for word-initial triphones

 

INFO: dict2pid.c(131): Allocated 3136 bytes (3 KiB) for word-final triphones

 

INFO: dict2pid.c(195): Allocated 3136 bytes (3 KiB) for single-phone word triphones

 

ERROR: "ngram_model_arpa.c", line 76: No \data\ mark in LM file

 

INFO: ngram_model_dmp.c(141): Will use memory-mapped I/O for LM file

 

INFO: ngram_model_dmp.c(195): ngrams 1=8, 2=10, 3=13

 

INFO: ngram_model_dmp.c(241):        8 = LM.unigrams(+trailer) read

 

INFO: ngram_model_dmp.c(289):       10 = LM.bigrams(+trailer) read

 

INFO: ngram_model_dmp.c(314):       13 = LM.trigrams read

 

INFO: ngram_model_dmp.c(338):        4 = LM.prob2 entries read

 

INFO: ngram_model_dmp.c(357):        5 = LM.bo_wt2 entries read

 

INFO: ngram_model_dmp.c(377):        3 = LM.prob3 entries read

 

INFO: ngram_model_dmp.c(405):        1 = LM.tseg_base entries read

 

INFO: ngram_model_dmp.c(461):        8 = ascii word strings read

 

INFO: ngram_search_fwdtree.c(99): 8 unique initial diphones

 

INFO: ngram_search_fwdtree.c(147): 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(186): Creating search tree

 

INFO: ngram_search_fwdtree.c(191): before: 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(324): after: max nonroot chan increased to 138

 

INFO: ngram_search_fwdtree.c(333): after: 5 root, 10 non-root channels, 3 single-phone words

 

INFO: ngram_search_fwdflat.c(153): fwdflat: min_ef_width = 4, max_sf_win = 25

 

INFO: continuous.c(261): ./pocketsphinx_continuous COMPILED ON: Feb 21 2011, AT: 22:31:47

 

 

 

READY....

 

Listening…

 

segment default….

 

 

 

1-adcde 设备选择中hw:AudioPCI pulseaudioalsa三者都试过,只有hw:AudioPCI可以成功。

 

2、向麦克风中说命令,发现出现segment default

 

三、第三次使用 

 

 1)重新录下五个.wav音频文件,使每个录音时间超过5s,保存为之前相同的名字。

 

2

 

./pocketsphinx_continuous -adcdev hw:AudioPCIhmm my_db.cd_cont_1000 -lm my_db.lm.DMP -dict my_db.dic

 

INFO: cmd_ln.c(512): Parsing command line:

 

./pocketsphinx_continuous \

 

       -adcdev hw:AudioPCI \

 

       -hmm my_db.cd_cont_1000 \

 

       -lm my_db.lm.DMP \

 

       -dict my_db.dic

 

 

 

Current configuration:

 

[NAME]        [DEFLT]        [VALUE]

 

-adcdev                        hw:AudioPCI

 

-agc        none              none

 

-agcthresh       2.0          2.000000e+00

 

-alpha            0.97        9.700000e-01

 

-argfile                 

 

-ascale            20.0        2.000000e+01

 

-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                      my_db.dic

 

-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                           my_db.cd_cont_1000

 

-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                       my_db.lm.DMP

 

-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                   

 

-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                    

 

-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(512): Parsing command line:

 

\

 

       -alpha 0.97 \

 

       -dither yes \

 

       -doublebw no \

 

       -nfilt 40 \

 

       -ncep 13 \

 

       -lowerf 133.33334 \

 

       -upperf 6855.4976 \

 

       -nfft 512 \

 

       -wlen 0.0256 \

 

       -transform legacy \

 

       -feat 1s_c_d_dd \

 

       -agc none \

 

       -cmn current \

 

       -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          8.0

 

-dither            no           yes

 

-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      1.333333e+02

 

-ncep             13           13

 

-nfft        512         512

 

-nfilt              40           40

 

-remove_dc    no           no

 

-round_filters  yes          yes

 

-samprate       16000            1.600000e+04

 

-seed              -1           -1

 

-smoothspec    no           no

 

-svspec                        

 

-transform      legacy            legacy

 

-unit_area       yes          yes

 

-upperf           6855.4976      6.855498e+03

 

-varnorm no           no

 

-verbose  no           no

 

-warp_params              

 

-warp_type     inverse_linear inverse_linear

 

-wlen             0.025625 2.560000e-02

 

 

 

INFO: acmod.c(238): Parsed model-specific feature parameters from my_db.cd_cont_1000/feat.params

 

INFO: fe_interface.c(288): You are using the internal mechanism to generate the seed.

 

INFO: feat.c(848): 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: mdef.c(520): Reading model definition: my_db.cd_cont_1000/mdef

 

INFO: bin_mdef.c(173): Allocating 166 * 8 bytes (1 KiB) for CD tree

 

INFO: tmat.c(205): Reading HMM transition probability matrices: my_db.cd_cont_1000/transition_matrices

 

INFO: acmod.c(117): Attempting to use SCHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 16644 variance values floored

 

INFO: acmod.c(119): Attempting to use PTHMM computation module

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/means

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(198): Reading mixture gaussian parameter: my_db.cd_cont_1000/variances

 

INFO: ms_gauden.c(292): 105 codebook, 1 feature, size

 

 8x39

 

INFO: ms_gauden.c(356): 16644 variance values floored

 

INFO: ptm_mgau.c(671): Reading mixture weights file 'my_db.cd_cont_1000/mixture_weights'

 

INFO: ptm_mgau.c(765): Read 105 x 1 x 8 mixture weights

 

INFO: ptm_mgau.c(831): Maximum top-N: 4

 

INFO: dict.c(294): Allocating 4104 * 20 bytes (80 KiB) for word entries

 

INFO: dict.c(306): Reading main dictionary: my_db.dic

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(309): 5 words read

 

INFO: dict.c(314): Reading filler dictionary: my_db.cd_cont_1000/noisedict

 

INFO: dict.c(206): Allocated 0 KiB for strings, 0 KiB for phones

 

INFO: dict.c(317): 3 words read

 

INFO: dict2pid.c(396): Building PID tables for dictionary

 

INFO: dict2pid.c(405): Allocating 16^3 * 2 bytes (8 KiB) for word-initial triphones

 

INFO: dict2pid.c(131): Allocated 3136 bytes (3 KiB) for word-final triphones

 

INFO: dict2pid.c(195): Allocated 3136 bytes (3 KiB) for single-phone word triphones

 

ERROR: "ngram_model_arpa.c", line 76: No \data\ mark in LM file

 

INFO: ngram_model_dmp.c(141): Will use memory-mapped I/O for LM file

 

INFO: ngram_model_dmp.c(195): ngrams 1=8, 2=10, 3=13

 

INFO: ngram_model_dmp.c(241):        8 = LM.unigrams(+trailer) read

 

INFO: ngram_model_dmp.c(289):       10 = LM.bigrams(+trailer) read

 

INFO: ngram_model_dmp.c(314):       13 = LM.trigrams read

 

INFO: ngram_model_dmp.c(338):        4 = LM.prob2 entries read

 

INFO: ngram_model_dmp.c(357):        5 = LM.bo_wt2 entries read

 

INFO: ngram_model_dmp.c(377):        3 = LM.prob3 entries read

 

INFO: ngram_model_dmp.c(405):        1 = LM.tseg_base entries read

 

INFO: ngram_model_dmp.c(461):        8 = ascii word strings read

 

INFO: ngram_search_fwdtree.c(99): 5 unique initial diphones

 

INFO: ngram_search_fwdtree.c(147): 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(186): Creating search tree

 

INFO: ngram_search_fwdtree.c(191): before: 0 root, 0 non-root channels, 4 single-phone words

 

INFO: ngram_search_fwdtree.c(324): after: max nonroot chan increased to 138

 

INFO: ngram_search_fwdtree.c(333): after: 5 root, 10 non-root channels, 3 single-phone words

 

INFO: ngram_search_fwdflat.c(153): fwdflat: min_ef_width = 4, max_sf_win = 25

 

INFO: continuous.c(261): ./pocketsphinx_continuous COMPILED ON: Feb 21 2011, AT: 22:31:47

 

 

 

READY....

 

Listening...

 

Stopped listening, please wait...

 

INFO: cmn_prior.c(121): cmn_prior_update: from <  8.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00  0.00 >

 

INFO: cmn_prior.c(139): cmn_prior_update: to   <  6.57 -0.33  0.07 -0.15 -0.02 -0.09  0.01 -0.15 -0.04 -0.06 -0.02 -0.06 -0.11 >

 

INFO: ngram_search_fwdtree.c(1513):      122 words recognized (2/fr)

 

INFO: ngram_search_fwdtree.c(1515):      534 senones evaluated (8/fr)

 

INFO: ngram_search_fwdtree.c(1517):      271 channels searched (4/fr), 59 1st, 151 last

 

INFO: ngram_search_fwdtree.c(1521):      151 words for which last channels evaluated (2/fr)

 

INFO: ngram_search_fwdtree.c(1524):        5 candidate words for entering last phone (0/fr)

 

INFO: ngram_search_fwdflat.c(295): Utterance vocabulary contains 1 words

 

INFO: ngram_search_fwdflat.c(912):        1 words recognized (0/fr)

 

INFO: ngram_search_fwdflat.c(914):      402 senones evaluated (6/fr)

 

INFO: ngram_search_fwdflat.c(916):      136 channels searched (2/fr)

 

INFO: ngram_search_fwdflat.c(918):       66 words searched (1/fr)

 

INFO: ngram_search_fwdflat.c(920):       48 word transitions (0/fr)

 

WARNING: "ngram_search.c", line 1087: </s> not found in last frame, using <s> instead

 

INFO: ngram_search.c(1137): lattice start node <s>.0 end node <s>.0

 

INFO: ps_lattice.c(1228): Normalizer P(O) = alpha(<s>:0:2) = -536874752

 

000000000: (null) (4427764)

 

READY....

 

Listening...

 

Stopped listening, please wait...

 

INFO: cmn_prior.c(121): cmn_prior_update: from <  6.57 -0.33  0.07 -0.15 -0.02 -0.09  0.01 -0.15 -0.04 -0.06 -0.02 -0.06 -0.11 >

 

INFO: cmn_prior.c(139): cmn_prior_update: to   <  6.59 -0.43  0.10  0.01  0.02 -0.07 -0.01 -0.13 -0.01 -0.09 -0.05 -0.10 -0.08 >

 

INFO: ngram_search_fwdtree.c(1513):       55 words recognized (1/fr)

 

INFO: ngram_search_fwdtree.c(1515):      489 senones evaluated (8/fr)

 

INFO: ngram_search_fwdtree.c(1517):      199 channels searched (3/fr), 33 1st, 97 last

 

INFO: ngram_search_fwdtree.c(1521):       97 words for which last channels evaluated (1/fr)

 

INFO: ngram_search_fwdtree.c(1524):       28 candidate words for entering last phone (0/fr)

 

INFO: ngram_search_fwdflat.c(295): Utterance vocabulary contains 1 words

 

INFO: ngram_search_fwdflat.c(912):       22 words recognized (0/fr)

 

INFO: ngram_search_fwdflat.c(914):      330 senones evaluated (5/fr)

 

INFO: ngram_search_fwdflat.c(916):      114 channels searched (1/fr)

 

INFO: ngram_search_fwdflat.c(918):       68 words searched (1/fr)

 

INFO: ngram_search_fwdflat.c(920):       31 word transitions (0/fr)

 

WARNING: "ngram_search.c", line 1087: </s> not found in last frame, using <sil> instead

 

INFO: ngram_search.c(1137): lattice start node <s>.0 end node <sil>.41

 

INFO: ps_lattice.c(1228): Normalizer P(O) = alpha(<sil>:41:59) = -79841

 

INFO: ps_lattice.c(1266): Joint P(O,S) = -79841 P(S|O) = 0

 

000000001: 右转 (-1415156)

 

READY....

 

 

 

 

 

 

 

 

 

主要参考网地:

 

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

 

2.       http://cmusphinx.sourceforge.net/wiki/faq

 

3.       http://ronaldramdhan.wordpress.com/2010/03/11/sphinxtrain/

 

4.       http://sourceforge.net/projects/cmusphinx/forums/forum/5471/topic/3939028

 

 

 

 

 

 

 

201132

 

展开阅读全文
打赏
0
1 收藏
分享
加载中
更多评论
打赏
0 评论
1 收藏
0
分享
返回顶部
顶部