Added basic Julius STT engine and vocabcompiler

[Holzhaus]

This adds a STT/Vocabulary implementation for the Open-Source Large Vocabulary CSR Engine Julius to Jasper.

Basically, recognition is working, but the freely available speaker-independent acoustic models offer a recognition rate that is probably too bad to use without custom training.

We COULD include this and state in the documentation that you need a custom trained acoustic model (you can define it via profile) and that we do not provide support for this step (and link to this tutorial).

@crm416 @shbhrsaha Do we want to merge this in on the above conditions?

[coveralls]

Coverage decreased (-1.74%) when pulling b969d37 on Holzhaus:julius-stt-engine into dd418d6 on jasperproject:master.

[shbhrsaha]

Yup, sounds good to give people the option. I support

[Holzhaus]

I added documentation for this here.

[shbhrsaha]

Not going to test this on the Pi, but I trust this worked for you on Arch so good to merge in

[andweber]

I tried the julius plugin and couldn't get it running out of the box.
It seems like the function definition of compile_vocabulary is not correct. That's why the profile options
"lexicon" and "lexicon_archive_member" are not recognized. I made a patch.

fix_def_julius_compile_vocabulary_patch.txt

Fixing this I'm stuck with the following error:

INFO:jasper.vocabcompiler:Starting compilation...
ERROR:jasper.vocabcompiler:Fatal compilation error occured
Traceback (most recent call last):
File "./jasper/vocabcompiler.py", line 140, in compile
compilation_func(config, self.path, phrases)
File "./plugins/stt/julius-stt/juliusvocab.py", line 133, in compile_vocabulary
subprocess.call(cmd, stdout=out_f, stderr=out_f)
File "/usr/lib/python2.7/subprocess.py", line 522, in call
return Popen(_popenargs, *_kwargs).wait()
File "/usr/lib/python2.7/subprocess.py", line 710, in init
errread, errwrite)
File "/usr/lib/python2.7/subprocess.py", line 1327, in _execute_child
raise child_exception
OSError: [Errno 2] No such file or directory

It seems like it is tried to call "mkdfa.pl". Is it missing??

edit
my fault sorry - was not able to install julius correctly

[andweber]

one more issue. "Jasper" as default keyword is dropped when compared with the dictionary because comparison is case sensitive.

I added two ways of fixing this:

1. always lookup with uppercase in translate_word
2. making sure that the keyword is always uppercase
   choose the best one.

fix_def_julius_compile_vocabulary_patch_3.txt

[Renezera]

Hey guys, I'm very new to all this and I'm having trouble setting up Julius-STT. Is there any step-by-step guide out there that can help me? Or is there any other STT that works better than the sphinx? I have some accent and sphinx doesn't recognizes 40% of the words I say lol.
Thanks again!