Voice Recognition Implementation

Naveen_D

@mrjj no...it is showing
ubuntu@ub:~/Documents/julius-4.2.2/SampleTest$ padsp julius -C julian.jconf
STAT: include config: julian.jconf
ERROR: Error in loading model
ubuntu@ub:~/Documents/julius-4.2.2/SampleTest$

mrjj

Ok , can mean many things.
http://voxforge.org/home/dev/acousticmodels/linux/create/htkjulius/tutorial/run-julian/comments/error-in-loading-model-when-executing-julius
Example
" the model tied state triphone you use is not contain sp."

So , you should check the modification you make and read docs on each command. etc to avoid this.

Naveen_D

@mrjj Now i want to use this julius in qt..how to use this..what i need to do?
can you please guide me in this matter..thanks

mrjj

@Naveen_D

Hi Always good starting reading docs.
https://julius.osdn.jp/juliusbook/en/desc_install.html

% make install
"which also installes headers and libraries as well as the binaries."

It means that it should generate a .so ? or .a file file and some .h files
( u must understand types of libs possible)
http://www.yolinux.com/TUTORIALS/LibraryArchives-StaticAndDynamic.html

you need to find out the exact names of those.

Then you need to add to a Qt project
http://doc.qt.io/qtcreator/creator-project-qmake-libraries.html

Make sure u fill it out correctly. Name wise.

Then you need find c++ example to see what functions to call etc.
and it should be it.

This might be ok sample
https://github.com/julius-speech/julius/blob/master/julius-simple/julius-simple.c

Naveen_D

@mrjj u said,
make install
"which also installes headers and libraries as well as the binaries."=
It means that it should generate a .so ? or .a file file and some .h files

Now I have configured and i run make and make install commands
After running sudo make install command, where i can get these .so or .a file in system (means which path) ?

Naveen_D

Hi i got the .a and .h files and binaries also...after this when i am trying with the example i need to add the library file and do i need to add all the .h files to my example project ?

mrjj

@Naveen_D
Hi
Normally you add a new path where to look for .h
files but you can also just include them.

like in the pro file
INCLUDEPATH += ../testLib
LIBS += -L../testLib/debug -ltestLib

https://wiki.qt.io/How_to_create_a_library_with_Qt_and_use_it_in_an_application

(NOTE , you names is NOT testLIB)

With the .A files
http://stackoverflow.com/questions/1361229/using-a-static-library-in-qt-creator

Naveen_D

@mrjj Hi i am bit confused after adding the library my .pro file is

#-------------------------------------------------
#
# Project created by QtCreator 2016-10-26T15:15:30
#
#-------------------------------------------------

QT       += core gui

greaterThan(QT_MAJOR_VERSION, 4): QT += widgets

TARGET = ExampleJulius
TEMPLATE = app


SOURCES += main.cpp\
        mywidget.cpp

HEADERS  += mywidget.h

unix:!macx: LIBS += -L$$PWD/./ -ljulius

INCLUDEPATH += $$PWD/.
DEPENDPATH += $$PWD/.

unix:!macx: PRE_TARGETDEPS += $$PWD/./libjulius.a

is it correct ?

jsulm

@Naveen_D

unix:!macx: LIBS += -L$$PWD/./ -ljulius
unix:!macx: PRE_TARGETDEPS += $$PWD/./libjulius.a

You should not link against shared lib and static lib (.a) at the same time. Either use shared lib or static lib.
Does $$PWD/. contain the lib itself and the header files?

Naveen_D

@jsulm I have copied the lib file and .h files to my example project and then in qt creator added the lib externally...

jsulm

@Naveen_D So, did you try to execute qmake and to do a complete rebuild to check whether it works?

Naveen_D

@jsulm Also i added the .h files to my project...
Ya i cleaned, run qmake and builded, it worked fine and executed

Naveen_D

@jsulm But whenever i opened the .h files in qt creator at top it is showing...
Error: Couldnot decode "callback.h" with "utf-8"-encoding. Editing not possible

mrjj

@Naveen_D

hi , you need to fix the file.

http://doc.qt.io/qtcreator/creator-editor-options-text.html

"To resolve the issue, use a file conversion tool such as Recode to convert the file encoding to UTF-8 when developing Qt 5 applications. Otherwise, conversion of string constants to QStrings might not work as expected."

http://superuser.com/questions/116907/how-to-recode-to-utf-8-conditionally

The "recode-to-utf8.sh " post.

Naveen_D

@mrjj Thanks...
Now i want know how to use the api's of julius..
I have developed a simple UI in qml in which i have developed simple icon with image and want to add voice recognition using julius...
Is it the rite way what i am doing now or there is any better way to achieve this?
how to do this..? guidance required..Thanks

mrjj

@Naveen_D
Hi

I have developed a simple UI in qml in which i have developed simple icon with image and want to add voice recognition
Is it the rite way what i am doing now or there is any better way to achieve this?

Hi, using QML for Interface is fine idea but why not wait a bit before having to learn how to call c++ functions from
QML and for now focus on getting the recognition up and running using plain c++ GUI app for testing.
Anyway, should not be super involving.
http://doc.qt.io/qt-5/qtqml-cppintegration-topic.html
http://stackoverflow.com/questions/9500280/access-c-function-from-qml

Then after you can call a c++ functions, its time to read the docs and study the sample.

Naveen_D

@mrjj Ya first i will try with the plain c++ GUI for voice recognition..then i will go for c++ and QML integration. Now how to proceed further with this voice recognition ?

Naveen_D

@mrjj Thanks alot for your advice

mrjj

@Naveen_D

Now how to proceed further with this voice recognition ?

Well I would starting with
https://github.com/julius-speech/julius/blob/master/julius-simple/julius-simple.c
Maybe make it your own class to wrap it up a bit or simply use as is.
Its pretty important you understand how this small sample works as
you need to be able to expand it for your application.
The main function accepts parameters and you must let the
app set it up and/or also allow user to change them.
So you most likely will split the main or change into something that fits your app.

UPDATE

What i ment - was that you need the same API calls as the samples uses

jconf = j_config_load_file_new(jconf_filename);
/* create recognition instance according to the jconf /
/ it loads models, setup final parameters, build lexicon
and set up work area for recognition */
recog = j_create_instance_from_jconf(jconf);
if (recog == NULL) { // ERRROR
}

/* register result callback functions */
callback_add(recog, CALLBACK_EVENT_SPEECH_READY, status_recready, NULL);
callback_add(recog, CALLBACK_EVENT_SPEECH_START, status_recstart, NULL);
callback_add(recog, CALLBACK_RESULT, output_result, NULL);

and more

Naveen_D

@mrjj Sorry i didn't get...I thought i have to make use of the api's that julius provide and i have to develop my own code..and wherever i want to make use of voice recognition i need to call the api provided by them.