Czech-speaking computer image lets you read his lips

14-12-2004

Computer generated human-like images that can talk are nothing extraordinary these days, but you would struggle to find one that can pronounce Czech words. Czech scientists have recently developed a synthetic human face that can pull off all the tongue-twisters of the Czech language and move his face accordingly. In today's Czech Science we meet "Chatter", as the image has been christened, and find out about the uses of such a device.

"Hello, I'm Chatter," says a 3D computer image, devised by researchers from the Technical University in the northern town of Liberec. In cooperation with the Institute of Radio-Electronics in Prague that took care of its synthetic voice, the team at the SpeechLab at the Technical University in Liberec developed an artificial male face that moves around as it pronounces the sounds.

"It has been shown by some researchers that you can better understand this synthetic speech coming from your computer if you see at the same time something like a face that mimics the movements of the lips and tongue. So the aim of this artificial talking face is to work together with the computer voice synthesis, text-to-speech system and help users of this synthetic speech to understand it better."

Professor Jan Nouza is the head of the research team at the SpeechLab in Liberec. He says that teaching Chatter Czech pronunciation is no easy task.

"Well, it was difficult and it is difficult, because we know it still needs a lot of improvement. We first started to work with a prototype of a system that was developed in the USA and we tried to redevelop their system of the speaking face for the Czech language and it showed that it didn't work so well. So that's why we tried to start with our own development of our own talking face."

Such talking images have more uses than just computer games. For example, they can help people with impaired hearing in a number of ways.

"In the future if the talking face is more natural than it is today, it really could help the people with impaired hearing, to assist them in understanding what is being said. Because these people can do what is called lip-tracking, such a talking face could help them in doing this lip-tracking for artificial speech, for example. There is also a possibility to make the model of the talking head transparent. So the user will not see the skin of the face but he or she can see through the skin how the vocal tract, how the mouth cavity, the lips, the teeth and tongue move when different words or different vowels and consonants are pronounced."

Find more information at http://itakura.kes.vslib.cz/kes/projekty.html.

14-12-2004