Voice Controlled Boe-Bot
I completed this project for my Senior Design project not too long ago. It is a voice controlled Parallax Boe-Bot which utilizes the Sphinx4 speech recognition engine. The speech recognition is decoded on my laptop in which the results are then sent to the Boe-Bot via Bluetooth. The speech recognition engine is coded in Java while the Boe-Bot in PBasic obviously.
The commands I have implemented include:
-"Move forward (1-9) feet"
-"Move backwards (1-9) feet"
-"Move left (1-9) feet"
-"Move right (1-9) feet"
-"Turn left (30/60/90) degrees"
-"Turn right (30/60/90) degrees"
-"Play a song" (piezospeaker melody)
-"Activate follow mode" (IR detection to follow objects)
-"Activate remote control" (Manual control with a Universal Remote)
Although Bluetooth Class2 had its range limitations, I was quite satisfied with the operating range. The speech recognition has a high accuracy, which would obviously decrease as more and more command phrases are added.
The video I have only shows a few of the commands. I will get around to making a full detailed video when I have some time.
Edit: Attached the .bs2 code. Basically where the Main begins, it awaits input over a RFCOMM port via the EB500. When my java app on PC recognizes a speech command, it will send a series of 3 variables to the Stamp. This comes in as [noparse][[/noparse]var1, var2, var3]. The Stamp code then proceeds to act accordingly depending on what variables were received.
To get started on the speech recognition, visit this site:
http://cmusphinx.sourceforge.net/sphinx4/
Follow through the tutorial and demos to gain a general understanding. You can then proceed to add your own phrases to be recognized through speech. I then used the RXTX java api to communicate via Bluetooth with the Stamp.
Post Edited (rotary9k) : 9/18/2008 5:25:51 PM GMT
The commands I have implemented include:
-"Move forward (1-9) feet"
-"Move backwards (1-9) feet"
-"Move left (1-9) feet"
-"Move right (1-9) feet"
-"Turn left (30/60/90) degrees"
-"Turn right (30/60/90) degrees"
-"Play a song" (piezospeaker melody)
-"Activate follow mode" (IR detection to follow objects)
-"Activate remote control" (Manual control with a Universal Remote)
Although Bluetooth Class2 had its range limitations, I was quite satisfied with the operating range. The speech recognition has a high accuracy, which would obviously decrease as more and more command phrases are added.
The video I have only shows a few of the commands. I will get around to making a full detailed video when I have some time.
Edit: Attached the .bs2 code. Basically where the Main begins, it awaits input over a RFCOMM port via the EB500. When my java app on PC recognizes a speech command, it will send a series of 3 variables to the Stamp. This comes in as [noparse][[/noparse]var1, var2, var3]. The Stamp code then proceeds to act accordingly depending on what variables were received.
To get started on the speech recognition, visit this site:
http://cmusphinx.sourceforge.net/sphinx4/
Follow through the tutorial and demos to gain a general understanding. You can then proceed to add your own phrases to be recognized through speech. I then used the RXTX java api to communicate via Bluetooth with the Stamp.
Post Edited (rotary9k) : 9/18/2008 5:25:51 PM GMT
bs2
10K
Comments
Could you attach your source code
i want to try
thanks for your help
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Engineering