[speech] Make ROS wrapper

ar13pit commented 6 years ago

Deadline: End of december 2018

speech_recognition is being built with a library that has no ROS bindings.

A new ROS wrapper needs to be written, preferably similar to the one being used with DragonFly to maintain code modularity.

ar13pit commented 5 years ago

Make a new repository called ros-kaldi and wrap the decoding scripts in C++.

ar13pit commented 5 years ago

@SamAlexandrov There are two things you need to do here

[ ] Identify the bash scripts/ C++ classes to wrap ROS around to run the recognition from a audio file.
[ ] Replace recognition from audio file by creating a audio stream from either a microphone or file, preferably in C++ (pretty much like any video processing methods)

The isolated digits scripts are enough for this.

PS: There may exist a streaming api in Kaldi, I haven't checked it. But if it doesn't then we will need to use an external library or create one ourselves.

PPS: I'll look into the grammar aspect as I have a feeling that I am close to finding the solution there.

tue-robotics / tue_robocup

[speech] Make ROS wrapper #666