VoxForge
VoxForge
Hello everybody, my problem is the following : I have french acoustic models for sphinx3 and I try integrate them in pocketSphinx.
I already try pocket Sphinx with english acoustic model and there was no problem.
My command is :
pocketsphinx_batch \
-adcin yes \
-dict /home/pocket-sphinx/pocketsphinx-0.5-fr/DESDHIS-2_sphinx.dico \
-hmm /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci \
-ctl /home/pocket-sphinx/pocketsphinx-0.5-fr/ca-fr.ctl \
-cepext .raw \
-cepdir /home/pocket-sphinx/pocketsphinx-0.5-fr/cepstra \
-fdict /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120.filler \
-lm /home/pocket-sphinx/pocketsphinx-0.5-fr/DESDHIS-2-Interpole.sphinx.trigram.DMP \
-wip 0.8 \
-lw 12 \
-hyp res.trn
And I have an error message :
INFO: mdef.c(520): Reading model definition: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/mdef
INFO: bin_mdef.c(157): cd_tree: nodes 176 wpos start 0 ci start 4 lc start 176 rc start 176
INFO: tmat.c(204): Reading HMM transition probability matrices: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/transition_matrices
INFO: acmod.c(108): Attempting to use SCGMM computation module
INFO: s2_semi_mgau.c(985): Reading S3 mixture gaussian file '/home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/means'
ERROR: "s2_semi_mgau.c", line 1015: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/means: #codebooks (129) != 1
INFO: acmod.c(121): Falling back to general multi-stream GMM computation
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/means
INFO: ms_gauden.c(292): 129 codebook, 1 feature, size
32x39
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/variances
INFO: ms_gauden.c(292): 129 codebook, 1 feature, size
32x39
INFO: ms_gauden.c(369): 0 variance values floored, 0 zero-variance components removed
INFO: ms_senone.c(157): Reading senone mixture weights: /home/pocket-sphinx/pocketsphinx-0.5-fr/bref120ci/mixture_weights
INFO: ms_senone.c(208): Truncating senone logs3(pdf) values by 10 bits
INFO: ms_senone.c(272): Read mixture weights for 129 senones: 1 features x 32 codewords
INFO: ms_mgau.c(124): The value of topn: 4
INFO: feat.c(849): 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: dict.c(232): Allocating 20 placeholders for new OOVs
ERROR: "dict.c", line 556: '=PLCHLDR0=': Unknown phone 'SIL'
ERROR: "dict.c", line 243: Failed to add DUMMY(SIL) entry to dictionary
FATAL_ERROR: "batch.c", line 381: PocketSphinx decoder init failed
If you have any suggestions , thank you very much.
PS : excuse me for english mistakes but I am French and I don't speak english very often.
--- (Edited on 7/24/2009 4:08 am [GMT-0500] by Visitor) ---
add -dictcase yes to the list of the arguments. use cmu sphinx forums and search there tpo get answers like this one.
--- (Edited on 7/24/2009 4:41 am [GMT-0500] by nsh) ---
thank you for your quick answer.
I have no longer the previous problem but I obviously a new problem :
For each word of my dictionnary, I have the following message :
WARNING: "ngram_search_fwdtree.c", line 115: Filler word 202 = accompagne(2) has more than one phone, ignoring it.Seemingly, Sphinx confuses the filler and the the dictionnary but I don't know why.
I am going to find an answer on the CMU Sphinx forum.
--- (Edited on 7/24/2009 7:01 am [GMT-0500] by Visitor) ---
Bonjour,
Did you succeed to migrate the french sphinx3 acoustic model on pocketsphinx?
Regards,
Laurent
--- (Edited on 5-October-2009 4:28 pm [GMT+0200] by laurent) ---
I'm wondering that i'm also facing same problem and adding the -dictbase 'yes' in etc/feat.params does not help me anymore. Would be greatful to get a reply.
Thank
--- (Edited on 6/1/2010 1:33 am [GMT-0500] by Visitor) ---
Same problem with simple following pocketsphinx_continuous command :
pocketsphinx_continuous \
-hmm /home/tom/Documents/build/sphinx/lium_french_f2/ \
-lm /home/tom/Documents/build/sphinx/french3g62K.lm.dmp \
-dict /home/tom/Documents/build/sphinx/frenchWords62K.dic \
-adcdev text.wav
...
dict.c(232): Allocating 20 placeholders for new OOVs
ERROR: "dict.c", line 556: '=PLCHLDR0=': Unknown phone 'SIL'
ERROR: "dict.c", line 243: Failed to add DUMMY(SIL) entry to dictionary
--- (Edited on 4/19/2012 4:46 pm [GMT-0500] by ) ---
I'm using lium files (http://www-lium.univ-lemans.fr/fr/content/ressources).
"means" file alway return an error by s2_semi_mgau.c :
"s2_semi_mgau.c", line 1011: /home/tom/Public/sphinx/lium_acoustic_models/parameters/F2/means: #codebooks (5725) != 1
Ressources files provided are to be used with sphinx 3.x, are compatible with pocketsphinx ?
--- (Edited on 4/21/2012 4:01 am [GMT-0500] by ) ---
Hello. LIUM models are compatible with pocketsphinx. Please use latest pocketsphinx version to avoid issues like described. Pocketsphinx-0.7 works just fine
Please download french acoustic models from cmusphinx website, not from the old location. Our website contains all the models we have in a single location. Models are kept up-to-date.
--- (Edited on 4/22/2012 14:31 [GMT+0400] by nsh) ---
Yes, thank you : ubuntu pocketsphinx version is quite old, sorry.
Is it possible to train pocketsphinx ? Does it lead to better results ?
--- (Edited on 4/26/2012 4:24 pm [GMT-0500] by tom) ---
> Is it possible to train pocketsphinx ?
Yes, see http://cmusphinx.sourceforge.net/wiki/tutorialadapt
> Does it lead to better results ?
Ye
--- (Edited on 4/27/2012 15:05 [GMT+0400] by nsh) ---