Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5

» Members: 893
» Latest member: Spyryt
» Forum threads: 601
» Forum posts: 3,191

Full Statistics

Controller and software
While waiting for my new 3D printer and inbetween filler and sanding the dome, I am doing some testing with my future controller system and the software to manage my R2 unit.

As I wrote in some other post, I will use some Raspberry Pi to control almost everything inside my robot. I know this board well enough to make it useful for this task. Now I am far from having a full controller as I am just testing small chunks of software. In the future these parts will be joined into one full system. I will put these code parts in my github repository for all the poeple interested on them...

You already saw a couple of small videos about the front logic displays and the holographic projector. Now I will show you my progress with speech recognition. I wish to control my R2 by voice as long as it's possible or at least have some kind of vocal interaction with it.

This is the video of a small test:

Note the following:
- I am using an headset and recording with a smartphone, so the audio is not the best
- I am italian and of course I speak italian!
- In Italy R2D2 was initially called C1P8 (I think it was a dubbing choice to better match labials). I am calling the robot this way, but it also respond to standard R2D2 name
- You can hear a dog barking outside my house. This shows also that the software is immune to some little noise

This is what's happening (translated in english):
I start saying "Hey C1". After the acknowledge beep (random from 4 choices) I continue by asking "what time is it" and the software waits for about 1 second of silence, then it tries to understand what I said.
When the software found what I am asking for it answers by saying "it's 17 and 35".
I called him again, this time with "C1P8". Again an acknowledge beep (this time a bit shorter and barely audible from the video) and I ask "what will the weather be tomorrow?". The answer is "Rain".

You can see some printed message from the software on the display (my tablet used as SSH terminal). These messages include also the answers from R2 (the time and the weather forecats).

Some tech notes:
- To recognize the name of the robot (hotwords) I'm using Snowboy. It's fast and works just locally (no need for internet!). Unfortunately it's not exactly speaker independent, even if it recognizes also my wife speech.
- To recognize the rest of the spoken sentences I use and this needs an internet connection. I tried PocketSphinx (it doesn't need a connection), but it's too bad at recognizing words, at least in italian
- The "voice" of R2 is built much like the r2d2translator website. For every letter/number there is a short sound. I create the final R2 "sentence" by joining the respective sounds from the text I wish to make it say (the printed answer in the messages of the display)

Hope you like it!  Smile

Forum Jump:

Users browsing this thread: 1 Guest(s)