Researchers at the Department of Engineering at the University of Cambridge and the Toshiba’s Cambridge Research Lab have come up with a new virtual assistant that may not be able to do any office work for you, but it sure would make your phones and computational devices a lot more human than they are now.
The program is called Zoe and uses the head of Hollyoaks star Zoe Lister. Based on the actress who plays Zoe Carpenter in the series airing on Channel 4, the talking head is able to replicate and express the full range of expression that humans use. Of course, the user has to input the emotion it wants Zoe to stimulate, though having a virtual talking head on your phone or tablet talking to you in tones that communicate fear, anger, happiness, etc. could be quite a wondrous thing indeed. We don’t really know how the program can be developed into becoming a digital personal assistant in the future, though having a computer simulated “person” not talking to us in a creepy robotic voice has got to be a pleasant start for fans of such technology.
As a user controllable avatar, Zoe is the most expressive one ever and it is precisely this unprecedented realism that sets it apart from other computer based emotional and voice simulation programs. The research team behind the program used thousands of sentences to make the system as realistic as possible and as a result, the speech model is much closer to the modulation that real life humans use. By using computer vision software, Zoe was also able to track the user’s face which helped it reflect the emotions on the user’s face too. To recreate expressions on a digital face, researchers used voice and image data, mathematical algorithms and voice and face-modeling to allow Zoe to decipher what expression to emote directly from the inputted text.
Zoe runs using a program that uses just tens of MB of data which would allow it to be installed and run smoothly on most current smartphones. For the average person, the program would allow functionality like customizable avatar that will allow them to use the image of their friends’ heads and their voices to read out texts. The emotionally realistic digital assistant would also make the otherwise bland digital assistant software used by professionals more human without the cost of hiring an actual human for the interaction. The template-based technology could also help children with autism and hearing disabilities learn human emotions and read lips. Researchers hope to have the technology make user interfaces, online lectures, audio-visual books, gaming etc. to make them more human friendly.