Sogou is introducing a new virtual news anchor who will help Chinese broadcasters communicate with the hearing impaired. The 3D digital avatar can be integrated into a telecast to translate the spoken content into sign language in real time.
The anchor has been given the name Xiao Cong, and is based on an earlier version of the AI news anchor already in Sogou’s Vocational Avatar portfolio. The new version was built with the China Common Sign Language Dictionary, and draws on several other Sogou technologies to make the end result feel more lifelike and realistic. For example, the hand gestures are realized with digital modeling, synthesis, and animation tech, while facial recognition was used to improve the quality of the avatar’s lip movements and facial expressions.
As a result, Sogou was able to deliver a product that better simulates the mannerisms of sign language as conveyed by a real human being. Xiao Cong can achieve an accuracy rate as high as 85 percent during a video broadcast, making it well-suited to public service announcements and educational programs in addition to standard news broadcasts.
Sogou is hoping that Xiao Cong will make telecasts more useful (and therefore more enjoyable) for China’s 27 million residents with hearing issues. The virtual news anchor was introduced at the 2021 Sohu Tech 5G & AI Conference on May 17.
“We are thrilled to bring the AI sign language news anchor to hearing impaired viewers,” said Sogou CEO Wang Xiaochuan. “We hope to help them communicate better and get more engaged with the society through our technological innovations.”
Sogou’s computer vision technology pushed the company to the top of the leaderboard in the most recent Facial Landmark Localization Grand Challenge. CyberLink and NTT DATA also used virtual avatars to communicate with shoppers at an unstaffed concept store in Tokyo.
May 18, 2021 – by Eric Weiss