Acoustic Model Discussions

Flat
Julius English grammar with lm_giga_64k_mvp_3gram
User: yuhan
Date: 6/1/2012 9:25 pm
Views: 3248
Rating: 7

Hi all,

I'm new to Julius and trying to set it up for English recognition. I following some discussion about the lm_giga_64_nvp_3gram is good for English, but got stucked with thousands lines of error showing "voca_load_htkdict: triphone not found".


Could someone point out the part that I did wrong? Thank you.

 

Here are my steps:

1. On Ubuntu 10.10, install Julius and voxforge acoustic model

      sudo apt-get install julius julius-voxforge

2. Download  lm_giga_64k_mvp_3gram.zip, and unpack it.

3. constructing the grammar file:

      mkbingram -nlr lm_giga_64k_nvp_3gram.arpa julius.bin

4. download a single channel wav file 10001-90210-01803.wav

5. Run julius: (provide input sound file, grammar (lm_giga) , dictionary (lm_giga), hmm (voxforge acoustic model from Ubuntu)

julius -input file -filelist ~/wav/10001-90210-01803.wav -d ./lm_giga_julius.bin -v ~/lm_giga_64k_nvp_3gram/lm_giga_64k_nvp.sphinx.dic -h /usr/share/julius-voxforge/acoustic/hmmdefs

 

Here are the error messsages that I got:

 

Error: voca_load_htkdict: the line content was: amritsar(2)    ae m r ah t s aa r
Error: voca_load_htkdict: line 2093: triphone "ae-m+r" not found
Error: voca_load_htkdict: line 2093: triphone "m-r+ow" not found
Error: voca_load_htkdict: the line content was: amro    ae m r ow
Error: voca_load_htkdict: line 2097: triphone "ae-m+t" not found
Error: voca_load_htkdict: line 2097: triphone "m-t+r" not found
Error: voca_load_htkdict: the line content was: amtrak    ae m t r ae k
Error: voca_load_htkdict: line 2098: triphone "ae-m+t" not found
Error: voca_load_htkdict: line 2098: triphone "m-t+r" not found
Error: voca_load_htkdict: the line content was: amtrak's    ae m t r ae k s
Error: voca_load_htkdict: line 2099: triphone "m-y+ah" not found
Error: voca_load_htkdict: line 2099: triphone "y-ah+l" not found
Error: voca_load_htkdict: line 2099: triphone "ah-l+ah" not found
Error: voca_load_htkdict: line 2099: triphone "l-ah+t" not found
Error: voca_load_htkdict: the line content was: amulets    ae m y ah l ah t s
Error: voca_load_htkdict: line 2100: triphone "ey-m+ah" not found
Error: voca_load_htkdict: line 2100: triphone "ah-n+d" not found
Error: voca_load_htkdict: the line content was: amundsen    ey m ah n d s ah n
Error: voca_load_htkdict: line 2101: triphone "aa-m+ah" not found
Error: voca_load_htkdict: line 2101: triphone "ah-n+d" not found
Error: voca_load_htkdict: the line content was: amundsen(2)    aa m ah n d s ah n
Error: voca_load_htkdict: line 2102: triphone "ae-m+ah" not found
Error: voca_load_htkdict: line 2102: triphone "ah-n+d" not found
Error: voca_load_htkdict: the line content was: amundson    ae m ah n d s ah n
Error: voca_load_htkdict: line 2103: triphone "ah-m+y" not found

 

--- (Edited on 6/1/2012 9:25 pm [GMT-0500] by yuhan) ---

PreviousNext