English broadcast news demo error

[vijayabharadwaj gsr]

I want to create broadcast news recognition demo. I have some files of MSNBC
news which are recorded at 16khz and 16bit.

I extracted mfcc using the following command

$SPHINXBASE/bin/sphinx_fe -c feats.ctl -samprate 16000 -upperf 8000 -lowerf 1
-remove_dc yes -nfilt 26 -transform dct -di wav -do feats -ei wav -eo mfc

Now if I use these features for recognition in Sphinx3, I am getting the
following error

INFO: kbcore.c(433): Begin Initialization of Core Models:
INFO: cmd_ln.c(512): Parsing command line:
\
-nfilt 26 \
-lowerf 1 \
-upperf 8000 \
-wlen 0.025 \
-transform dct \
-round_filters no \
-remove_dc yes \
-feat 1s_c_d_dd \
-svspec 0-12/13-25/26-38 \
-agc none \
-cmn current \
-varnorm no

Current configuration:

-agc none none
-agcthresh 2.0 2.000000e+00
-alpha 0.97 9.700000e-01
-ceplen 13 13
-cmn current current
-cmninit 8.0 8.0
-dither no no
-doublebw no no
-feat 1s_c_d_dd 1s_c_d_dd
-frate 100 100
-input_endian little little
-lda
-ldadim 0 0
-lifter 0 0
-logspec no no
-lowerf 133.33334 1.000000e+00
-ncep 13 13
-nfft 512 512
-nfilt 40 26
-remove_dc no yes
-round_filters yes no
-samprate 16000 1.600000e+04
-seed -1 -1
-smoothspec no no
-svspec 0-12/13-25/26-38
-transform legacy dct
-unit_area yes yes
-upperf 6855.4976 8.000000e+03
-varnorm no no
-verbose no no
-warp_params
-warp_type inverse_linear inverse_linear
-wlen 0.025625 2.500000e-02

INFO: Initialization of the log add table
INFO: Log-Add table size = 29350 x 2 >> 0
INFO:
INFO: feat.c(848): Initializing feature stream to type: '1s_c_d_dd',
ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean= 12.00, mean= 0.0
INFO: kbcore.c(480): .cont.
INFO: Initialization of feat_t, report:
INFO: Feature type = 1s_c_d_dd
INFO: Cepstral size = 13
INFO: Number of streams = 1
INFO: Vector size of stream: 39
INFO: Number of subvectors = 0
INFO: Whether CMN is used = 1
INFO: Whether AGC is used = 0
INFO: Whether variance is normalized = 0
INFO:
INFO: Reading HMM in Sphinx 3 Model format
INFO: Model Definition File: models/hmm//mdef
INFO: Mean File: models/hmm//means
INFO: Variance File: models/hmm//variances
INFO: Mixture Weight File: models/hmm//mixture_weights
INFO: Transition Matrices File: models/hmm//transition_matrices
INFO: mdef.c(682): Reading model definition: models/hmm//mdef
INFO: Initialization of mdef_t, report:
INFO: 50 CI-phone, 142729 CD-phone, 3 emitstate/phone, 150 CI-sen, 5150 Sen,
26579 Sen-Seq
INFO:
INFO: kbcore.c(288): Using optimized GMM computation for Continuous HMM, -topn
will be ignored
INFO: cont_mgau.c(163): Reading mixture gaussian file 'models/hmm//means'
FATAL_ERROR: "cont_mgau.c", line 206: #Features streams(3) != 1 in continuous
HMM

I could decode these files using HUB4 models. What modifications I need to do
to create this demo using English broadcast news acoustic models.

please help me.

[Nickolay V. Shmyrev]

I extracted mfcc using the following command $SPHINXBASE/bin/sphinx_fe -c
feats.ctl -samprate 16000 -upperf 8000 -lowerf 1 -remove_dc yes -nfilt 26
-transform dct -di wav -do feats -ei wav -eo mfc

I'm not sure why you've done this but it doesn't look like you are doing
proper thing or understand what you are doing.

FATAL_ERROR: "cont_mgau.c", line 206: #Features streams(3) != 1 in
continuous HMM

This error was caused by -svspec 0-12/13-25/26-38 which was taken from
feat.params from your model. Again, since you didn't manage to provide full
log, it's impossible to say how did this feat.params went into sphinx3 decoder
setup.

[vijayabharadwaj gsr]

I did the feature extraction by looking at feat.params distributed with
English broadcast models.

-nfilt 26
-lowerf 1
-upperf 8000
-wlen 0.025
-transform dct
-round_filters no
-remove_dc yes
-feat 1s_c_d_dd
-svspec 0-12/13-25/26-38
-agc none
-cmn current
-varnorm no

[vijayabharadwaj gsr]

Actually I am using the model downloaded from the Sphinx site.

[Nickolay V. Shmyrev]

I did the feature extraction by looking at feat.params distributed with
English broadcast models.

They are for pocketsphinx, they are not compatible with sphinx3.

[vijayabharadwaj gsr]

I am sorry If I am foolish. Are you saying feature extraction values I have
used are for pocket sphinx or the English broadcast news models distributed in
sphinx site are for pocket sphinx.

[Nickolay V. Shmyrev]

English broadcast news models distributed in sphinx site are for pocket
sphinx.

Models are for pocketsphinx