This is the bower package for the HUB4 acoustic model to be used with pocketsphinx.js.
HUB4 is an acoustic model for US English, trained on broadcast news audio. It is available as part of PocketSphinx. Is should be used with audio recorded at 8kHz.
This bower package depends on pocketsphinx.js-lib which contains the library, without any language-specific file, such as acoustic or language model.
For general usage see the docs of the base library (once installed, in bower_components/pocketsphinx.js-lib/README.md).
Following are the specific for that model.
a. Loading the acoustic model
The files must be loaded in the recognizer, with a call to the "load" command prior to initialization. The following file names should be used:
data = {command: 'load', data: ["feat.params.js", "mdef.js", "means.js", "noisedict.js", "sendump.js", "transition_matrices.js", "variances.js"].map(function(x) {return "../pocketsphinx.js-en_US-hub4wsj_sc_8k/" + x;}b. Initializing the recognizer
At init time, this model can be selected with:
["-hmm", "hub4wsj_sc_8k"]c. Setting the sample rate
The recorder should be initialized with the correct sample rate, 8kHz:
outputSampleRate: 8000These files are compiled versions of the acoustic models shipped with PocketSphinx source code. PocketSphinx licensing terms are included in the pocketsphinx and sphinxbase folders of PocketSphinx.js source repository.