NEC is developing a new gestural speech software with Semantics Technology Group that will give robots an extra layer of realism when speaking. The software generates synthetic gestures such as head nodding or tilting which are performed while the robot speaks and not when the robot is silent. Normally these gestures would be programmed manually, so if the software proves successful it could cut time and costs substantially. Although PaPeRo has fewer axes of rotation in its neck than other humanoid robots, for the time being the software will be tested on PaPeRo before adding further complexity.
NEC is also developing a more natural speech synthesizer database to improve human-robot interactions. Current text-to-speech software reads written news articles, announcements, emails and so on in a monotonous style that isn’t suitable for intelligent robots. The new software uses a conversational two-stage audio recording process, and preliminary studies suggest it is much more convincing compared to earlier solutions.
[source: Robonable (JP)]