 brian
|
I had my app running English voice recognition, switched some stuff to do Chinese recognition, then switched it back for English configuration. Well, I thought I switched everything back, but now I’m getting this error and I’m not sure what is causing it. I complete redownloaded and replaced my “models” file for pocketsphinx to be sure nothing I did in there might be affecting it.
Here’s my log readings where it says one of the the errors is occurring:
” INFO: ms_gauden.c(358): 0 variance values floored
INFO: s2_semi_mgau.c(897): Loading senones from dump file /var/mobile/Applications/4412F7DD-DA40-4C23-A2E3-6E74724C6855/SmallTalk.app/sendump
INFO: s2_semi_mgau.c(921): BEGIN FILE FORMAT DESCRIPTION
ERROR: “s2_semi_mgau.c”, line 998: Number of senones mismatch: 5210 != 5150
INFO: acmod.c(119): Attempting to use PTHMM computation module
INFO: ms_gauden.c(198): Reading mixture gaussian parameter: /var/mobile/Applications/4412F7DD-DA40-4C23-A2E3-6E74724C6855/SmallTalk.app/means
INFO: ms_gauden.c(292): 1 codebook, 3 feature, size…”
It runs through a couple of error similar to the one above and the ends with the following:
“…INFO: ms_senone.c(224): Transposing mixture weights in memory
ERROR: “ms_senone.c”, line 265: Weight normalization failed for 3 senones
INFO: ms_senone.c(277): Read mixture weights for 6144 senones: 1 features x 8 codewords
INFO: ms_senone.c(318): Mapping all senones to one codebook
FATAL_ERROR: “ms_mgau.c”, line 110: #Feature mismatch: gauden= 3, senone= 1
Debugger stopped.”
Any help on what needs fixed? This seemed to be stemming from files in hub4wsj_sc_8k, but after reloading it, nothing changed.
|
 Halle
|
The way that OpenEars works currently, you can only have a single acoustic model folder added to the bundle — the other must be completely removed. Is the old one completely removed?
|
 brian
|
Yes.
I found some extra files left over from some copy I had put in in my project’s folder, but they hadn’t been added to the project, so taking them out didn’t make a difference.
There’s is only one acoustic model in the project.
|
 Halle
|
Hmm, wish I had some advice for you but it really looks like something is left over from the wrong model in spite of your steps. Have you done all the stuff to make sure nothing is cached like deleting your app from your device, throwing out your build folders and the contents of DerivedData, throwing out your acoustic model contents and your lm and phonetic dictionary and doing a search for any related files in your project that might be hiding in a group somewhere, and then re-adding the correct items back in and then finally compiling?
|
 brian
|
Oh, I cleaned the build targets and it fixed that error. Should have thought of that. Thanks!
|