Servo - May 2010

VRbot Module

THE VRBOT GUI.

YOU’LL HAVE TO ENUNCIATE MORE IF YOU WANT GENTLEMEN TO THINK YOU HAD A PROPER UPBRINGING.

up our commands, even at a conversational volume. We pushed our luck by trying to whisper and seeing if the bot could hear. Understandably, it preferred us to speak up, but even at a decent distance the microphone could pick up a loud whisper.

The real fun comes with programming custom voice commands, and the fairly intuitive process is once again detailed in the instruction video and in a PDF file on the Tigal website. After connecting the robot to the computer, simply select the command group that you want to edit. Using the buttons on the toolbar, you can choose to add a command to the group. After naming the command something descriptive, the next step is to train it. Training comes in two phases where the user must simply say the command. After training, the test group function is a great way to see that your efforts worked.

You can even give your robot a name that it responds to by programming a new trigger word. We, of course, chose to train the name that we had already given the Robonova, though we decided against pretentiousness and dropped the title so that we would simply call the bot by “Myshkin.” As we were messing around with different voice commands, we wondered how the VRbot module would deal with similar sounding words. Fortunately, the developers have thought of that very problem, and we tested their solution in true Pygmalion fashion. First, we trained the command “rain,” and we predictably did so without incident. Next was the command “Spain,” which we also trained without a problem. Finally, we trained the word “plain,” and this brought up a warning. A dialog box cautioned us that this was similar to one of our other trained commands. The command was flagged with a yellow sign, and we proceeded to test the group with some trepidation. With careful diction, we recited the apparent truism about Spanish weather, and we were pleased to see that all the words were recognized correctly. The next time through we let ourselves get a little sloppy, and the test was unable to distinguish between the words “rain” and “plain.” Trying again with our best Queen’s English, the program recognized all of the words. So, it seems that users can train their Robonova in limericks if they so desire, but they must simply be prepared to speak clearly. Professor Higgins would be proud.