pocketsphinx_batch wav file decoding quality

[Roberts]

I am trying to set up pocketsphinx to recognize speech from wav files. Recognition results are a lot worse, compared to results from sphinxtrain acoustic model evaluation results.

If I use feature file instead of wav file, than the results are the same, so I assume, that the problem might be related to configuration of feature extraction, but I can't find it. Any suggestions?

So here are all the files:

etc/sphinx_train.cfg http://pastebin.com/DwcmiNbm

etc/feat.params

    -nfilt __CFG_NUM_FILT__
    -lowerf __CFG_LO_FILT__
    -upperf __CFG_HI_FILT__
    -feat __CFG_FEATURE__
    -svspec __CFG_SVSPEC__
    -agc __CFG_AGC__
    -cmn __CFG_CMN__
    -varnorm __CFG_VARNORM__

command I use for pocketsphinx_batch

pocketsphinx_batch ^
    -hmm "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/files_20130709_tr_tr_filtred/model_parameters/all.cd_cont_200" ^
    -dict "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/files_20130709_tr_tr_filtred/etc/all.dic" ^
    -lm "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/files_20130709_tr_tr_filtred/etc/all.lm.DMP" ^
    -ctl "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/audio/test.txt" ^
    -cepdir "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/audio" ^
    -cepext .wav ^
    -adcin true
    -hyp batch_result.txt ^
    -lw 10 ^
    -beam 1e-80 ^
    -wip 0.2 ^
    -wbeam 1e-40 ^
    -ctloffset 0 ^
    -ctlcount 2 ^
    -nfilt 40 ^
    -lowerf 133.3334 ^
    -upperf 6855.4976 ^
    -feat 1s_c_d_dd ^
    -agc none ^
    -cmn current ^
    -varnorm no ^
    -nbestdir "D:/Desktop/Desktop/haltura/LUMII/runa/sphinx/pocketsphinx-0.8-win32/test/audio" ^
    -nbest 10000

[Nickolay V. Shmyrev]

Share the files you are trying to decode and the model training folder.