0

When I use pocketsphinx_continuous -infile my audio -dict my_dictionary -jsgf my_jsgf

I got these errors:

ERROR: "dict.c", line 195: Line 1: Phone 'I' is mising in the acoustic model; word 'Bismi' ignored

ERROR: "dict.c", line 195: Line 2: Phone 'I' is mising in the acoustic model; word 'Bismi(2)' ignored

ERROR: "dict.c", line 195: Line 3: Phone 'A' is mising in the acoustic model; word 'Lahi' ignored

ERROR: "dict.c", line 195: Line 4: Phone 'A' is mising in the acoustic model; word 'Lahi(2)' ignored

ERROR: "dict.c", line 195: Line 5: Phone 'HI' is mising in the acoustic model; word 'Rahmani' ignored

ERROR: "dict.c", line 195: Line 6: Phone 'HI' is mising in the acoustic model; w

ERROR: "fsg_search.c", line 141: The word 'wa' is missing in the dictionary

But, I got the a good accuracy when I tested my data in the result file

TOTAL Words: 420 Correct: 420 Errors: 0

TOTAL Percent correct = 100.00% Error = 0.00% Accuracy = 100.00%

TOTAL Insertions: 0 Deletions: 0 Substitutions: 0

How can i solve theses problems? Thank you so much

Community
  • 1
  • 1
hasan
  • 17
  • 3

1 Answers1

0

You need to edit the dictionary to match the acoustic model phoneset or use the acoustic model to match your dictionary. Acoustic model is configured with -hmm option.

Nikolay Shmyrev
  • 24,897
  • 5
  • 43
  • 87
  • i used -hmm option i didn't get the text i got these INFO: continuous.c(305): pocketsphinx_continuous COMPILED ON: Mar 8 2016, AT: 23:45:54 INFO: cmn_prior.c(131): 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(149): cmn_prior_update: to < 57.55 15.24 -2.83 2.28 -13.93 -4.90 16.14 -10.01 18.20 -2.65 10.41 1.24 10.51 > INFO: fsg_search.c(843): 310 frames, 2347 HMMs (7/fr), 6071 senones (19/fr), 926 history entries (2/fr) How can i solve this problem?Thank you – hasan Mar 11 '16 at 14:52
  • To see just the text you can add an option to disable logs: `-logfn /dev/null` – Nikolay Shmyrev Mar 11 '16 at 15:00
  • i got nothing . what could be my mistake ? hasan@hasan-SVE14112EGW:~/Desktop/as/sphinxtrain$ pocketsphinx_continuous -infile '/home/hasan/Desktop/as/sphinxtrain/an4/wav/an4test_clstk/td/an1-td-b.wav' -hmm '/home/hasan/Desktop/as/sphinxtrain/model_parameters' -dict '/home/hasan/Desktop/as/sphinxtrain/etc/an4.dic' -jsgf '/home/hasan/Desktop/as/sphinxtrain/etc/an4.jsgf' -logfn /dev/null – hasan Mar 11 '16 at 15:21
  • There is no mistake, nothing is also a result. Probably your audio is too short, try longer audio. To recognize short audio add a line `-cmninit 57,15,-2` to model feat.params file – Nikolay Shmyrev Mar 11 '16 at 15:35
  • Thank you so much .You help me a lot .Thank you my lecture – hasan Mar 11 '16 at 15:51