Two Japanese tinker with a human voice apparatus
It doesn’t look like that, the speaking apparatus of Hideyuki Sawada. Nor should it. Sawada from Kagawa University in Japan is concerned with the message, and it should sound as human as possible. Let others worry about the humanoid Hulle. New Scientist magazine reports that Sawada, in collaboration with Shuji Hashimoto of Waseda University, has built a speech apparatus that he plans to use to equip humanoid robots.
So far, the vocal tract consists of a thick-walled flesh-colored silicone tube that is squeezed from below by individually movable presses. The apparatus is ventilated by a lung-like structure. A kind of glottis and vocal cords are also present. Teeth, tongue, palate, larynx and the like are still missing, but Sawada and Hashimoto will come up with something. In the meantime, a neural network makes the sounds more and more realistic.
Sawada plans to present the device at the International Conference on Robotics and Automation in Washington next week.C. Prasent. If you don’t have the time or the inclination to jet off to Washington, you can take a look at the current repertoire of the speech apparatus on the web: At first glance, both experimental setups look the same, except for the image section. The first example deals with vowels, the second with consonants. To be honest, the sounds are more reminiscent of the screaming unit in a teddy bar jerk than of the human voice. Sometimes the sounds are lost in the stomping of the hydraulics.
Quite absurd, if you consider that the human voice can be produced synthetically quite convincing. So well, in fact, that developers like Juergen Schroeter of AT&T Laboratories in New Jersey are already thinking about possible misuse and the incorporation of an acoustic watermark.
On the other hand, AI researcher Luc Steels takes a very similar approach in his Talking Heads experiment (cf. Werthern pushes Lotte) takes a very similar approach. He also equips newer generations of his speech robots with a naturalistic vocal tract, which supposedly already sound quite human-like.
Probably the enthusiasm for mechanics can be explained only historically. The 18. And 19. Century was obsessed with the idea of lifelike automata: the writing dolls of Maillardet and Jaquet-Droz, von Kempelen’s chess automaton (later exposed as a fake), and finally – even if only as fiction – E.T.A. Hoffmann’s machine-woman Olympia, who turns the head of the student Nathanael. They were all invented to test the boundaries between man and machine. Not to forget Vaucanson’s duck, which allegedly had a functioning digestive tract. This legend celebrated its resurrection last year at the Kunsthalle Wien. There, Belgian conceptual artist Wim Delvoye presented his "Cloaca", an oversized digestive system that actually turned food into discs (cf. Shit happens). Against this background it was high time to deal with the human vocal tract. At some point the circle must close. Until Sawada’s apparatus is ready for the museum, however, a few more years were allowed to pass by.