5.3.2 CI Phone Models: Single-Path Silence Training
Silence training of phone models is similar to silence training of word models. Both processes use similar parameters and the same transcription database. The only difference is in the language model, which now has an extra level for the phonetic contents of each word. See Section 4.2.4 for more information on the phone level of the language model file. The details of single-path silence training are explained in Section 5.2.3. Go to the directory:
Command: isip_recognize -param params_sil.sof -verbose brief -list $ISIP_TUTORIAL/research/isip/databases/lists/identifiers_train.sof Version: 1.23 (not released) 2003/05/21 23:10:45 loading audio database: $ISIP_TUTORIAL/research/isip/databases/db/tidigits_audio_db.sof *** no symbol graph database file was specified *** loading transcription database: $ISIP_TUTORIAL/research/isip/databases/db/tidigits_trans_word_db.sof loading front-end: $ISIP_TUTORIAL/recipes/frontend.sof loading language model: $ISIP_TUTORIAL/models/ci_phone_models/lm_phone_jsgf_init.sof loading statistical model pool: $ISIP_TUTORIAL/models/ci_phone_models/smp_phone_init.sof loading configuration file: $ISIP_TUTORIAL/sections/s05/s05_03_p03/config.sof starting iteration: 0 processing file 1 (ae_12a): $ISIP_TUTORIAL/research/isip/databases/sof_8k/train/ae_12a.sof retrieving annotation graph for identifier: ae_12a, level: word transcription: ONE TWO average utterance probability: -70.621869668053364, number of frames: 110 processing file 2 (ae_1a): $ISIP_TUTORIAL/research/isip/databases/sof_8k/train/ae_1a.sof retrieving annotation graph for identifier: ae_1a, level: word transcription: ONE ....As with word models, the next step is multi-path silence training. Continue to the next section to begin. |
