I saw this singing robot at Wired magazine. Watch the video and read the subtitles closely, since much of it is in Japanese. This is Cool Science! A robot diva programmed to mimic a human singer’s facial movements breathing patterns.
Japanese engineers have taken a different tack in their robotic development, working to make more lifelike robots that will integrate well into human society. To do this, they’ve followed some interesting learning curves. I liked how they use maps of mouth movement and even integrating breathing patterns into the robot’s speech.
Did you catch the reference to using hidden Markov technology to train the robot when to breath? What they are referring to is the Markov (“mark-off”) probability model, often used in statistical modeling. All the more reason to make sure that our kids know their math. Math is key to technological progress, and we need our kids to be eager to learn it. I think videos like this help build the desire to learn and they help eliminate any negative stigma about math that their math-phobic friends and teachers may be inadvertently passing on to them.
Here’s what goes on in Hidden Markov Modeling, or Hidden Markov Speech Recognition. The machines calculate the correlations in speech patterns that take place, nuances which we might not pay attention to, but which are in fact highly correlated to how we will speak. For example, you’re talking away and the robot or computer is taking this all down, and calculating correlations between what you said sixty seconds ago with what you said later at 59, 58, 57, and 56 or less seconds later. This way, the computer/robot begins to calculate highly effective forecasts with what you will say next, since it has been calculating the probabilities from what you’ve said before.
If you’re interested in the history of Japanese robots and the new directions in which they are headed, here is a great book which I highly recommend: Loving the Machine: The Art and Science of Japanese Robots by Timothy N Hornyak, which I have reviewed before.