//THE PROGRAM IS WORKING FINE WITH MEANS VARIANCE AND ALL FROM CI_CONT MODELS AS SHOWN BELOW UNTIL // CTRL-C IS PRESSED //I have modified sphinx to print the values of hyp->sf and hyp->ef which is creating problem //when cd_cont models are used shyam@karanattu:~/tutorial/sharika$ ./sharika-sh sphinx2-continuous sphinx2-continuous \ -fsgfn ./sharika.fsg \ -mdeffn ./model_parameters/sharika.ci_cont/mdef \ -meanfn ./model_parameters/sharika.ci_cont/means \ -varfn ./model_parameters/sharika.ci_cont/variances \ -mixwfn ./model_parameters/sharika.ci_cont/mixture_weights \ -dictfn ./etc/sharika.dic \ -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models \ -hmmext chmm karanattu വ്യാ ഒക്ടോ 18 05:01:53 IST 2007 INFO: fbs_main.c(1304): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: logs3.c(107): Initializing logbase: 1.000100e+00 INFO: logs3.c(168): Log-Add table size = 99042 INFO: mdef.c(605): Reading model definition: ./model_parameters/sharika.ci_cont/mdef INFO: mdef.c(775): 9 CI-phone, 0 CD-phone, 5 emitstate/phone, 45 CI-sen, 45 Sen, 9 Sen-Seq INFO: s3mdef_s2map.c(94): Writing /tmp/__s3_s2__.phone INFO: s3mdef_s2map.c(127): Writing /tmp/__s3_s2__.map INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.ci_cont/means' INFO: cont_mgau.c(229): 45 mixture Gaussians, 1 components, veclen 39 INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.ci_cont/variances' INFO: cont_mgau.c(229): 45 mixture Gaussians, 1 components, veclen 39 INFO: cont_mgau.c(250): Reading mixture weights file './model_parameters/sharika.ci_cont/mixture_weights' INFO: cont_mgau.c(336): Read 45 x 1 mixture weights INFO: cont_mgau.c(352): Removing uninitialized Gaussian densities INFO: cont_mgau.c(382): 0 densities removed (0 mixtures removed entirely) INFO: cont_mgau.c(390): Applying variance floor INFO: cont_mgau.c(404): 0 variance values floored INFO: cont_mgau.c(448): Precomputing Mahalanobis distance invariants INFO: kb_main.c(601): Reading phone file [/tmp/__s3_s2__.phone] INFO: phone.c(387): Using 0 phonological mappings INFO: kb_main.c(610): Reading map file [/tmp/__s3_s2__.map] INFO: hmm_tied_r.c(1342): Read Map: 9 phones map to 9 unique senone sequences INFO: kb_main.c(614): Checking mdef and phone/map files for consistency INFO: kb_main.c(648): Reading dict file [./etc/sharika.dic] INFO: dict.c(483): First line of ./etc/sharika.dic was not '!'; will assume no header present INFO: dict.c(667): 8 = words in file [./etc/sharika.dic] INFO: dict.c(409): LEFT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 0 triphones 0 pseudo diphones 72 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 0 triphones 0 pseudo diphones 72 uniphones INFO: dict.c(1248): 8 right context entries INFO: dict.c(1249): 1 ave entries per exit context INFO: dict.c(413): RIGHT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 0 triphones 0 pseudo diphones 72 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 0 triphones 0 pseudo diphones 72 uniphones INFO: dict.c(1248): 8 right context entries INFO: dict.c(1249): 1 ave entries per exit context INFO: dict.c(417): 0 unique triphones were mapped to ci phones INFO: search.c(3661): Estimating maximal search tree INFO: search.c(3772): max nonroot chan increased to 128 INFO: search.c(3782): 0 root, 0 non-root channels, 11 single-phone words INFO: search.c(2054): topsen-window = 1, no phone-prediction INFO: search.c(3056): -1105296 = beam width INFO: search.c(3065): -648968 = new word beam width INFO: search.c(3083): -1105296 = new phone beam width INFO: search.c(3092): -921080 = last phone beam width INFO: search.c(3074): -648968 = Last phone alone beam width INFO: search.c(3136): -52986 = LOG (Silence Word Penalty) + LOG (Phone Penalty) INFO: search.c(3149): -184216 = LOG (Filler Word Penalty) + LOG (Phone Penalty) INFO: search.c(3127): 0 = newword penalty INFO: search.c(3160): LW = fwdtree: 6.5, fwdflat: 8.5, bestpath: 9.5 INFO: search.c(3162): Log2(AscrScale) = 0 INFO: search.c(4217): Flat-pass bw = 1.0e-08 (-1473728), nwbw = 3.0e-04 (-648968) INFO: fe_sigproc.c(746): Current FE Parameters: INFO: fe_sigproc.c(747): Sampling Rate: 16000.000000 INFO: fe_sigproc.c(748): Frame Size: 410 INFO: fe_sigproc.c(749): Frame Shift: 160 INFO: fe_sigproc.c(750): FFT Size: 512 INFO: fe_sigproc.c(751): Lower Frequency: 133.333 INFO: fe_sigproc.c(752): Upper Frequency: 6855.5 INFO: fe_sigproc.c(753): Number of filters: 40 INFO: fe_sigproc.c(754): Number of Overflow Samps: 0 INFO: fe_sigproc.c(755): Start Utt Status: 0 INFO: fe_sigproc.c(760): Will not add dither to audio INFO: fe_sigproc.c(765): Will not use double bandwidth in mel filter INFO: live_norm.c(123): mean[0]= 8.00, mean[1..12]= 0.0 INFO: fsg_search.c(248): FSG(beam: -1105296, pbeam: -1105296, wbeam: -648968; wip: -28002, pip: 0) INFO: word_fsg.c(798): Reading FSG file './sharika.fsg' (altpron=1, filler=1, lw=6.50, silprob=5.00e-03, fillprob=1.00e-08) INFO: word_fsg.c(340): Adding filler words to FSG INFO: word_fsg.c(612): FSG: 11 states, 29 transitions (10 null, 0 alt, 11 filler, 0 unknown) INFO: word_fsg.c(287): Computing transitive closure for null transitions INFO: word_fsg.c(321): 17 null transitions added INFO: fsg_lextree.c(115): 19 HMM nodes in lextree INFO: uttproc.c(1827): FSG= "sc" INFO: uttproc.c(1142): FSG Mode; lextree, flat, bestpath searches disabled INFO: uttproc.c(1885): CMN: Based on current utterance INFO: uttproc.c(1905): AGC: None INFO: fbs_main.c(1441): /* Need to select from among multiple FSGs */ INFO: search.c(2144): startword= (id= 9) INFO: fbs_main.c(1451): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: tty-continuous.c(221): sphinx2-continuous COMPILED ON: Feb 24 2006, AT: 12:14:05 READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000001) (4/8/16/30/32K): 90.1%(18682) 5.5%(1147) 3.7%(772) 0.7%(135) 0.0%(0); max: 25157 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-3559914) > best final state score (-3715790); but using latter sf:127 ef:127 sf:26 ef:127 sf:0 ef:25 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000001) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 25 -1127967 -344409 -43383 1 1 -1.00 SIL 26 127 -1931662 0 -18937 1 5 0.57 PADU 127 127 0 -155876 0 5 10 -1.00 ---------------------------------------------------------------------------------------- 0 127 -3059629 -656161 10F 00000001(TOTAL) FSGSRCH: PADU (00000001 -3715790 (A=-3059629 L=-656161)) INFO: fsg_search.c(1142): Utt 00000001: 128 frames, 1735 HMMs (13/fr), 5760 senones (45/fr), 1671 history entries (13/fr) INFO: uttproc.c(558): 1.28 SoS, 2.19 sec elapsed, 1.71 xRT, 0.00 sec CPU, 0.00 xRT 128: PADU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000002) (4/8/16/30/32K): 82.9%(21439) 9.8%(2534) 6.3%(1624) 1.0%(259) 0.0%(0); max: 27793 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-4578536) > best final state score (-4734412); but using latter sf:159 ef:159 sf:108 ef:159 sf:87 ef:107 sf:0 ef:86 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000002) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 86 -2544594 -344409 -29248 1 1 -1.00 SIL 87 107 -511051 -344409 -24335 1 1 -1.00 SIL 108 159 -678197 0 -13042 1 5 0.97 PADU 159 159 0 -155876 0 5 10 -1.00 ---------------------------------------------------------------------------------------- 0 159 -3733842 -1000570 10F 00000002(TOTAL) FSGSRCH: PADU (00000002 -4734412 (A=-3733842 L=-1000570)) INFO: fsg_search.c(1142): Utt 00000002: 160 frames, 2597 HMMs (16/fr), 7200 senones (45/fr), 2728 history entries (17/fr) INFO: uttproc.c(558): 1.60 SoS, 2.93 sec elapsed, 1.83 xRT, 0.00 sec CPU, 0.00 xRT 160: PADU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000003) (4/8/16/30/32K): 87.5%(53060) 6.4%(3911) 4.9%(2971) 1.2%(729) 0.0%(1); max: 31248 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-14772020) > best final state score (-14927896); but using latter sf:376 ef:376 sf:112 ef:376 sf:54 ef:111 sf:53 ef:53 sf:28 ef:53 sf:0 ef:27 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000003) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 27 -1154713 -344409 -41239 1 1 -1.00 SIL 28 53 -856579 0 -32945 1 5 0.89 PADU 53 53 0 -467628 0 5 1 -1.00 54 111 -1816851 0 -31325 1 4 0.49 ADUTHATHU 112 376 -9631555 -344409 -36345 4 4 -1.00 SIL 376 376 0 -155876 0 4 10 -1.00 ---------------------------------------------------------------------------------------- 0 376 -13459698 -1468198 10F 00000003(TOTAL) FSGSRCH: PADU ADUTHATHU (00000003 -14927896 (A=-13459698 L=-1468198)) INFO: fsg_search.c(1142): Utt 00000003: 377 frames, 6158 HMMs (16/fr), 16965 senones (45/fr), 5446 history entries (14/fr) INFO: uttproc.c(558): 3.77 SoS, 5.25 sec elapsed, 1.39 xRT, 0.01 sec CPU, 0.00 xRT 377: PADU ADUTHATHU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000004) (4/8/16/30/32K): 85.7%(11631) 9.3%(1261) 4.3%(581) 0.7%(93) 0.0%(2); max: 31551 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-3480148) > best final state score (-3636024); but using latter sf:82 ef:82 sf:21 ef:82 sf:0 ef:20 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000004) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 20 -863999 -344409 -41142 1 1 -1.00 SIL 21 82 -2115864 0 -34126 1 7 0.30 EMACS 82 82 0 -155876 0 7 10 -1.00 ---------------------------------------------------------------------------------------- 0 82 -2979863 -656161 10F 00000004(TOTAL) FSGSRCH: EMACS (00000004 -3636024 (A=-2979863 L=-656161)) INFO: fsg_search.c(1142): Utt 00000004: 83 frames, 1290 HMMs (15/fr), 3735 senones (45/fr), 1290 history entries (15/fr) INFO: uttproc.c(558): 0.83 SoS, 1.65 sec elapsed, 1.99 xRT, 0.00 sec CPU, 0.00 xRT 83: EMACS READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000005) (4/8/16/30/32K): 92.9%(12843) 2.9%(396) 3.1%(434) 1.1%(150) 0.0%(1); max: 32256 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-3729362) > best final state score (-3953085); but using latter sf:83 ef:83 sf:56 ef:83 sf:0 ef:55 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000005) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 55 -2181414 -344409 -38953 1 1 -1.00 SIL 56 83 -1115510 0 -39839 1 3 0.26 CHURUKKU 83 83 0 -155876 0 3 10 -1.00 ---------------------------------------------------------------------------------------- 0 83 -3296924 -656161 10F 00000005(TOTAL) FSGSRCH: CHURUKKU (00000005 -3953085 (A=-3296924 L=-656161)) INFO: fsg_search.c(1142): Utt 00000005: 84 frames, 1538 HMMs (18/fr), 3780 senones (45/fr), 2092 history entries (24/fr) INFO: uttproc.c(558): 0.84 SoS, 1.63 sec elapsed, 1.94 xRT, 0.00 sec CPU, 0.00 xRT 84: CHURUKKU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000006) (4/8/16/30/32K): 89.3%(28571) 5.6%(1794) 4.1%(1310) 1.0%(325) 0.0%(0); max: 30423 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-5657754) > best final state score (-6011010); but using latter sf:197 ef:197 sf:12 ef:197 sf:0 ef:11 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000006) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 11 -572222 0 -47685 1 3 0.26 CHURUKKU 12 197 -4782627 -344409 -25713 3 3 -1.00 SIL 197 197 0 -155876 0 3 10 -1.00 ---------------------------------------------------------------------------------------- 0 197 -5354849 -656161 10F 00000006(TOTAL) FSGSRCH: CHURUKKU (00000006 -6011010 (A=-5354849 L=-656161)) INFO: fsg_search.c(1142): Utt 00000006: 198 frames, 3536 HMMs (17/fr), 8910 senones (45/fr), 5937 history entries (29/fr) INFO: uttproc.c(558): 1.98 SoS, 2.72 sec elapsed, 1.37 xRT, 0.01 sec CPU, 0.00 xRT 198: CHURUKKU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000007) (4/8/16/30/32K): 71.3%(7118) 17.0%(1697) 9.6%(961) 2.1%(208) 0.0%(0); max: 30673 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (-1954861) > best final state score (-2110737); but using latter sf:59 ef:59 sf:10 ef:59 sf:0 ef:9 sf:-1 ef:-1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000007) ---------------------------------------------------------------------------------------- -1 -1 0 -155876 0 0 1 -1.00 0 9 -571450 -344409 -57145 1 1 -1.00 SIL 10 59 -883126 0 -17662 1 9 0.97 EAZHUTHIDAM 59 59 0 -155876 0 9 10 -1.00 ---------------------------------------------------------------------------------------- 0 59 -1454576 -656161 10F 00000007(TOTAL) FSGSRCH: EAZHUTHIDAM (00000007 -2110737 (A=-1454576 L=-656161)) INFO: fsg_search.c(1142): Utt 00000007: 60 frames, 785 HMMs (13/fr), 2700 senones (45/fr), 516 history entries (8/fr) INFO: uttproc.c(558): 0.60 SoS, 1.33 sec elapsed, 2.22 xRT, 0.00 sec CPU, 0.00 xRT 60: EAZHUTHIDAM READY.... C-c C-cINFO: uttproc.c(601): INFO: uttproc.c(603): TOTAL Elapsed time 17.70 seconds INFO: uttproc.c(605): TOTAL CPU time 0.02 seconds INFO: uttproc.c(607): TOTAL Speech 10.90 seconds INFO: uttproc.c(610): AVERAGE 1.62 xRT(Elapsed), 0.00 xRT(CPU) END ////For the above run accuracy is 100% and it shows what i have said..Here i quited manually by ////pressing ctrl-c..Now i tried sphinx2-continuous with cd_cont models whose output log is ////shown below..But below using the cd_cont it runs only for the first utterance and that itself //// is not detecting the right word.. shyam@karanattu:~/tutorial/sharika$ ./sharika2-sh sphinx2-continuous bash: ./sharika2-sh: Permission denied shyam@karanattu:~/tutorial/sharika$ su Password: karanattu:/home/shyam/tutorial/sharika# chmod a+x sharika2-sh karanattu:/home/shyam/tutorial/sharika# exit exit shyam@karanattu:~/tutorial/sharika$ ./sharika2-sh sphinx2-continuous sphinx2-continuous \ -fsgfn ./sharika.fsg \ -mdeffn ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef \ -meanfn ./model_parameters/sharika.cd_cont_1000_8.s2models/means \ -varfn ./model_parameters/sharika.cd_cont_1000_8.s2models/variances \ -mixwfn ./model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights \ -dictfn ./etc/sharika.dic \ -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models \ -hmmext chmm karanattu വ്യാ ഒക്ടോ 18 05:04:27 IST 2007 INFO: fbs_main.c(1304): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: logs3.c(107): Initializing logbase: 1.000100e+00 INFO: logs3.c(168): Log-Add table size = 99042 INFO: mdef.c(605): Reading model definition: ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef INFO: mdef.c(775): 9 CI-phone, 648 CD-phone, 5 emitstate/phone, 45 CI-sen, 370 Sen, 177 Sen-Seq INFO: s3mdef_s2map.c(94): Writing /tmp/__s3_s2__.phone INFO: s3mdef_s2map.c(127): Writing /tmp/__s3_s2__.map INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.cd_cont_1000_8.s2models/means' INFO: cont_mgau.c(229): 370 mixture Gaussians, 8 components, veclen 39 INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.cd_cont_1000_8.s2models/variances' INFO: cont_mgau.c(229): 370 mixture Gaussians, 8 components, veclen 39 INFO: cont_mgau.c(250): Reading mixture weights file './model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights' INFO: cont_mgau.c(336): Read 370 x 8 mixture weights INFO: cont_mgau.c(352): Removing uninitialized Gaussian densities 62 63 72 79 81 82 83 84 108 109 112 116 117 118 119 120 121 123 124 125 147 148 149 150 151 152 153 154 156 158 159 160 161 164 166 169 187 189 207 209 226 230 233 234 235 237 238 239 240 241 243 246 248 272 273 274 276 277 297 301 302 304 305 306 308 309 311 313 315 316 317 319 320 333 334 346 351 352 359 361 365 368 369 INFO: cont_mgau.c(382): 1403 densities removed (83 mixtures removed entirely) INFO: cont_mgau.c(390): Applying variance floor INFO: cont_mgau.c(404): 2276 variance values floored INFO: cont_mgau.c(448): Precomputing Mahalanobis distance invariants INFO: kb_main.c(601): Reading phone file [/tmp/__s3_s2__.phone] INFO: phone.c(387): Using 0 phonological mappings INFO: kb_main.c(610): Reading map file [/tmp/__s3_s2__.map] INFO: hmm_tied_r.c(1342): Read Map: 657 phones map to 177 unique senone sequences INFO: kb_main.c(614): Checking mdef and phone/map files for consistency INFO: kb_main.c(648): Reading dict file [./etc/sharika.dic] INFO: dict.c(483): First line of ./etc/sharika.dic was not '!'; will assume no header present INFO: dict.c(667): 8 = words in file [./etc/sharika.dic] INFO: dict.c(409): LEFT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1248): 32 right context entries INFO: dict.c(1249): 4 ave entries per exit context INFO: dict.c(413): RIGHT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1248): 38 right context entries INFO: dict.c(1249): 4 ave entries per exit context INFO: dict.c(417): 0 unique triphones were mapped to ci phones INFO: search.c(3661): Estimating maximal search tree INFO: search.c(3772): max nonroot chan increased to 128 INFO: search.c(3782): 0 root, 0 non-root channels, 11 single-phone words INFO: search.c(2054): topsen-window = 1, no phone-prediction INFO: search.c(3056): -1105296 = beam width INFO: search.c(3065): -648968 = new word beam width INFO: search.c(3083): -1105296 = new phone beam width INFO: search.c(3092): -921080 = last phone beam width INFO: search.c(3074): -648968 = Last phone alone beam width INFO: search.c(3136): -52986 = LOG (Silence Word Penalty) + LOG (Phone Penalty) INFO: search.c(3149): -184216 = LOG (Filler Word Penalty) + LOG (Phone Penalty) INFO: search.c(3127): 0 = newword penalty INFO: search.c(3160): LW = fwdtree: 6.5, fwdflat: 8.5, bestpath: 9.5 INFO: search.c(3162): Log2(AscrScale) = 0 INFO: search.c(4217): Flat-pass bw = 1.0e-08 (-1473728), nwbw = 3.0e-04 (-648968) INFO: fe_sigproc.c(746): Current FE Parameters: INFO: fe_sigproc.c(747): Sampling Rate: 16000.000000 INFO: fe_sigproc.c(748): Frame Size: 410 INFO: fe_sigproc.c(749): Frame Shift: 160 INFO: fe_sigproc.c(750): FFT Size: 512 INFO: fe_sigproc.c(751): Lower Frequency: 133.333 INFO: fe_sigproc.c(752): Upper Frequency: 6855.5 INFO: fe_sigproc.c(753): Number of filters: 40 INFO: fe_sigproc.c(754): Number of Overflow Samps: 0 INFO: fe_sigproc.c(755): Start Utt Status: 0 INFO: fe_sigproc.c(760): Will not add dither to audio INFO: fe_sigproc.c(765): Will not use double bandwidth in mel filter INFO: live_norm.c(123): mean[0]= 8.00, mean[1..12]= 0.0 INFO: fsg_search.c(248): FSG(beam: -1105296, pbeam: -1105296, wbeam: -648968; wip: -28002, pip: 0) INFO: word_fsg.c(798): Reading FSG file './sharika.fsg' (altpron=1, filler=1, lw=6.50, silprob=5.00e-03, fillprob=1.00e-08) INFO: word_fsg.c(340): Adding filler words to FSG INFO: word_fsg.c(612): FSG: 11 states, 29 transitions (10 null, 0 alt, 11 filler, 0 unknown) INFO: word_fsg.c(287): Computing transitive closure for null transitions INFO: word_fsg.c(321): 17 null transitions added INFO: fsg_lextree.c(115): 43 HMM nodes in lextree INFO: uttproc.c(1827): FSG= "sc" INFO: uttproc.c(1142): FSG Mode; lextree, flat, bestpath searches disabled INFO: uttproc.c(1885): CMN: Based on current utterance INFO: uttproc.c(1905): AGC: None INFO: fbs_main.c(1441): /* Need to select from among multiple FSGs */ INFO: search.c(2144): startword= (id= 9) INFO: fbs_main.c(1451): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: tty-continuous.c(221): sphinx2-continuous COMPILED ON: Feb 24 2006, AT: 12:14:05 READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000001) (4/8/16/30/32K): 90.7%(29489) 5.6%(1808) 3.3%(1063) 0.5%(152) 0.0%(0); max: 25309 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (1875482074) > best final state score (1875326198); but using latter sf:200 ef:200 sf:68 ef:200 sf:12 ef:67 sf:6 ef:11 sf:5 ef:5 sf:2 ef:5 sf:0 ef:1 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000001) ---------------------------------------------------------------------------------------- 0 1 1878701199 0 939350599 1 6 -1.00 VALAPARATHU 2 5 -416184 -344409 -104046 6 6 -1.00 SIL 5 5 0 -467628 0 6 1 -1.00 6 11 1125756 0 187626 1 6 -1.00 VALAPARATHU 12 67 -518719 -344409 -9262 6 6 -1.00 SIL 68 200 -1909123 -344409 -14354 6 6 -1.00 SIL 200 200 0 -155876 0 6 10 -1.00 ---------------------------------------------------------------------------------------- 0 200 1876982929 -1656731 10F 00000001(TOTAL) FSGSRCH: VALAPARATHU VALAPARATHU (00000001 1875326198 (A=1876982929 L=-1656731)) INFO: fsg_search.c(1142): Utt 00000001: 201 frames, 2655 HMMs (13/fr), 9279 senones (46/fr), 1491 history entries (7/fr) INFO: uttproc.c(558): 2.01 SoS, 2.86 sec elapsed, 1.42 xRT, 0.01 sec CPU, 0.00 xRT 201: VALAPARATHU VALAPARATHU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000002) (4/8/16/30/32K): 78.0%(9186) 13.7%(1613) 7.6%(891) 0.7%(86) 0.0%(0); max: 28762 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (1877023252) > best final state score (1876867376); but using latter sf:71 ef:71 sf:63 ef:71 sf:4 ef:62 sf:11 ef:3 sphinx2-continuous: fsg_history.c:294: fsg_history_entry_hyp_extract: Assertion `hyp->sf <= hyp->ef' failed. ./sharika2-sh: line 2: 12204 Aborted $1 -fsgfn ./sharika.fsg -mdeffn ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef -meanfn ./model_parameters/sharika.cd_cont_1000_8.s2models/means -varfn ./model_parameters/sharika.cd_cont_1000_8.s2models/variances -mixwfn ./model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights -dictfn ./etc/sharika.dic -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models -hmmext chmm END ////Finally i tried running the cd_cont models with gdb and the output is shown below.. shyam@karanattu:~/tutorial/sharika$ gdb GNU gdb 6.4.90-debian Copyright (C) 2006 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i486-linux-gnu". (gdb) file sphinx2-continuous Reading symbols from /usr/bin/sphinx2-continuous...(no debugging symbols found)...done. Using host libthread_db library "/lib/i686/cmov/libthread_db.so.1". (gdb) run -fsgfn ./sharika.fsg -mdeffn ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef -meanfn ./model_parameters/sharika.cd_cont_1000_8.s2models/means -varfn ./model_parameters/sharika.cd_cont_1000_8.s2models/variances -mixwfn ./model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights -dictfn ./etc/sharika.dic -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models -hmmext chmm Starting program: /usr/bin/sphinx2-continuous -fsgfn ./sharika.fsg -mdeffn ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef -meanfn ./model_parameters/sharika.cd_cont_1000_8.s2models/means -varfn ./model_parameters/sharika.cd_cont_1000_8.s2models/variances -mixwfn ./model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights -dictfn ./etc/sharika.dic -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models -hmmext chmm Failed to read a valid object file image from memory. (no debugging symbols found) /usr/bin/sphinx2-continuous \ -fsgfn ./sharika.fsg \ -mdeffn ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef \ -meanfn ./model_parameters/sharika.cd_cont_1000_8.s2models/means \ -varfn ./model_parameters/sharika.cd_cont_1000_8.s2models/variances \ -mixwfn ./model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights \ -dictfn ./etc/sharika.dic \ -hmmdirlist ./model_parameters/sharika.cd_cont_1000_8.s2models \ -hmmext chmm karanattu വ്യാ ഒക്ടോ 18 05:05:34 IST 2007 INFO: fbs_main.c(1304): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: logs3.c(107): Initializing logbase: 1.000100e+00 INFO: logs3.c(168): Log-Add table size = 99042 INFO: mdef.c(605): Reading model definition: ./model_parameters/sharika.cd_cont_1000_8.s2models/mdef INFO: mdef.c(775): 9 CI-phone, 648 CD-phone, 5 emitstate/phone, 45 CI-sen, 370 Sen, 177 Sen-Seq INFO: s3mdef_s2map.c(94): Writing /tmp/__s3_s2__.phone INFO: s3mdef_s2map.c(127): Writing /tmp/__s3_s2__.map INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.cd_cont_1000_8.s2models/means' INFO: cont_mgau.c(229): 370 mixture Gaussians, 8 components, veclen 39 INFO: cont_mgau.c(104): Reading mixture gaussian file './model_parameters/sharika.cd_cont_1000_8.s2models/variances' INFO: cont_mgau.c(229): 370 mixture Gaussians, 8 components, veclen 39 INFO: cont_mgau.c(250): Reading mixture weights file './model_parameters/sharika.cd_cont_1000_8.s2models/mixture_weights' INFO: cont_mgau.c(336): Read 370 x 8 mixture weights INFO: cont_mgau.c(352): Removing uninitialized Gaussian densities 62 63 72 79 81 82 83 84 108 109 112 116 117 118 119 120 121 123 124 125 147 148 149 150 151 152 153 154 156 158 159 160 161 164 166 169 187 189 207 209 226 230 233 234 235 237 238 239 240 241 243 246 248 272 273 274 276 277 297 301 302 304 305 306 308 309 311 313 315 316 317 319 320 333 334 346 351 352 359 361 365 368 369 INFO: cont_mgau.c(382): 1403 densities removed (83 mixtures removed entirely) INFO: cont_mgau.c(390): Applying variance floor INFO: cont_mgau.c(404): 2276 variance values floored INFO: cont_mgau.c(448): Precomputing Mahalanobis distance invariants INFO: kb_main.c(601): Reading phone file [/tmp/__s3_s2__.phone] INFO: phone.c(387): Using 0 phonological mappings INFO: kb_main.c(610): Reading map file [/tmp/__s3_s2__.map] INFO: hmm_tied_r.c(1342): Read Map: 657 phones map to 177 unique senone sequences INFO: kb_main.c(614): Checking mdef and phone/map files for consistency INFO: kb_main.c(648): Reading dict file [./etc/sharika.dic] INFO: dict.c(483): First line of ./etc/sharika.dic was not '!'; will assume no header present INFO: dict.c(667): 8 = words in file [./etc/sharika.dic] INFO: dict.c(409): LEFT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1248): 32 right context entries INFO: dict.c(1249): 4 ave entries per exit context INFO: dict.c(413): RIGHT CONTEXT TABLES INFO: dict.c(1119): Entry Context table contains 8 entries INFO: dict.c(1120): 72 possible cross word triphones. INFO: dict.c(1156): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1182): Exit Context table contains 8 entries INFO: dict.c(1183): 72 possible cross word triphones. INFO: dict.c(1246): 72 triphones 0 pseudo diphones 0 uniphones INFO: dict.c(1248): 38 right context entries INFO: dict.c(1249): 4 ave entries per exit context INFO: dict.c(417): 0 unique triphones were mapped to ci phones INFO: search.c(3661): Estimating maximal search tree INFO: search.c(3772): max nonroot chan increased to 128 INFO: search.c(3782): 0 root, 0 non-root channels, 11 single-phone words INFO: search.c(2054): topsen-window = 1, no phone-prediction INFO: search.c(3056): -1105296 = beam width INFO: search.c(3065): -648968 = new word beam width INFO: search.c(3083): -1105296 = new phone beam width INFO: search.c(3092): -921080 = last phone beam width INFO: search.c(3074): -648968 = Last phone alone beam width INFO: search.c(3136): -52986 = LOG (Silence Word Penalty) + LOG (Phone Penalty) INFO: search.c(3149): -184216 = LOG (Filler Word Penalty) + LOG (Phone Penalty) INFO: search.c(3127): 0 = newword penalty INFO: search.c(3160): LW = fwdtree: 6.5, fwdflat: 8.5, bestpath: 9.5 INFO: search.c(3162): Log2(AscrScale) = 0 INFO: search.c(4217): Flat-pass bw = 1.0e-08 (-1473728), nwbw = 3.0e-04 (-648968) INFO: fe_sigproc.c(746): Current FE Parameters: INFO: fe_sigproc.c(747): Sampling Rate: 16000.000000 INFO: fe_sigproc.c(748): Frame Size: 410 INFO: fe_sigproc.c(749): Frame Shift: 160 INFO: fe_sigproc.c(750): FFT Size: 512 INFO: fe_sigproc.c(751): Lower Frequency: 133.333 INFO: fe_sigproc.c(752): Upper Frequency: 6855.5 INFO: fe_sigproc.c(753): Number of filters: 40 INFO: fe_sigproc.c(754): Number of Overflow Samps: 0 INFO: fe_sigproc.c(755): Start Utt Status: 0 INFO: fe_sigproc.c(760): Will not add dither to audio INFO: fe_sigproc.c(765): Will not use double bandwidth in mel filter INFO: live_norm.c(123): mean[0]= 8.00, mean[1..12]= 0.0 INFO: fsg_search.c(248): FSG(beam: -1105296, pbeam: -1105296, wbeam: -648968; wip: -28002, pip: 0) INFO: word_fsg.c(798): Reading FSG file './sharika.fsg' (altpron=1, filler=1, lw=6.50, silprob=5.00e-03, fillprob=1.00e-08) INFO: word_fsg.c(340): Adding filler words to FSG INFO: word_fsg.c(612): FSG: 11 states, 29 transitions (10 null, 0 alt, 11 filler, 0 unknown) INFO: word_fsg.c(287): Computing transitive closure for null transitions INFO: word_fsg.c(321): 17 null transitions added INFO: fsg_lextree.c(115): 43 HMM nodes in lextree INFO: uttproc.c(1827): FSG= "sc" INFO: uttproc.c(1142): FSG Mode; lextree, flat, bestpath searches disabled INFO: uttproc.c(1885): CMN: Based on current utterance INFO: uttproc.c(1905): AGC: None INFO: fbs_main.c(1441): /* Need to select from among multiple FSGs */ INFO: search.c(2144): startword= (id= 9) INFO: fbs_main.c(1451): libfbs/main COMPILED ON: Oct 16 2007, AT: 03:55:36 INFO: tty-continuous.c(221): /usr/bin/sphinx2-continuous COMPILED ON: Feb 24 2006, AT: 12:14:05 READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000001) (4/8/16/30/32K): 93.5%(37083) 3.8%(1504) 2.4%(963) 0.3%(130) 0.0%(0); max: 27490 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (1875134534) > best final state score (1874978658); but using latter sf:245 ef:245 sf:63 ef:245 sf:7 ef:62 sf:0 ef:6 SFrm EFrm AScr LScr TSDiffPF SrcSt DstSt Conf Word (FSG) (00000001) ---------------------------------------------------------------------------------------- 0 6 1878816847 0 268402406 1 6 -1.00 VALAPARATHU 7 62 -479181 -344409 -8556 6 6 -1.00 SIL 63 245 -2514314 -344409 -13739 6 6 -1.00 SIL 245 245 0 -155876 0 6 10 -1.00 ---------------------------------------------------------------------------------------- 0 245 1875823352 -844694 10F 00000001(TOTAL) FSGSRCH: VALAPARATHU (00000001 1874978658 (A=1875823352 L=-844694)) INFO: fsg_search.c(1142): Utt 00000001: 246 frames, 3222 HMMs (13/fr), 11222 senones (45/fr), 1811 history entries (7/fr) INFO: uttproc.c(558): 2.46 SoS, 3.43 sec elapsed, 1.39 xRT, 0.01 sec CPU, 0.00 xRT 246: VALAPARATHU READY.... INFO: uttproc.c(1220): Batchmode Listening... Stopped listening, please wait... INFO: uttproc.c(1426): Samples histogram (00000002) (4/8/16/30/32K): 68.1%(8025) 15.3%(1802) 12.6%(1484) 3.9%(457) 0.1%(8); max: 32320 WARNING: "uttproc.c", line 790: NO AGC INFO: fsg_search.c(1036): Best score (1877333411) > best final state score (1877177535); but using latter sf:71 ef:71 sf:12 ef:71 sf:7 ef:11 sf:6 ef:6 sf:2 ef:6 Program received signal SIGSEGV, Segmentation fault. fsg_history_entry_hyp_extract (h=0x8071c30, id=2, hyp=0x81056a0) at fsg_history.c:289 289 hyp->sf = pred_entry->frame + 1; (gdb) q The program is running. Exit anyway? (y or n) y shyam@karanattu:~/tutorial/sharika$