Li Shiyan, director of Baidu’s Intelligent Cloud AI human-computer interaction lab, talks about the digital human sector in January. China Daily
The 2022 Winter Olympics have just come to an end, but the feast of ice and snow is still going on – Beijing Winter Paralympics kicked off on March 4. A sign language AI TV anchor that has served the Winter Olympics is now doing the broadcasting job for the Winter Paralympics, ensuring Chinese audiences who have hearing difficulties can enjoy the Games.
北京2022年冬奥会刚刚结束，但冰雪盛宴仍在继续——3月4日，北京冬残奥会拉开帷幕。一位曾在北京冬奥会上提供服务的AI（人工智能）手语主播，现在正在北京冬残奥会上提供手语广播服务，以确保听力有障碍的中国观众能够感受赛事精彩。
Created by Baidu Smart Cloud, the anchor is supported by the world’s largest sign language database with up to 200,000 pieces of data. Her mission is to provide a great service to those hearing-impaired audience, allowing them to quickly obtain event information.
冬奥AI手语主播由百度智能云创建，拥有世界上最大的手语数据库支持，内含多达20万条语料。AI手语主播的使命是为听障人士提供良好的服务，令其快速获得赛事信息。
The technology behind the virtual anchor ensures it can do the job. Sign language’s word order is different from spoken Mandarin. For example, people with hearing difficulties place the predicate at the end of the sentence, while we put them in the middle of the sentence, just between the subject and the object in modern Chinese. Therefore, sign language recognition is difficult, according to Yuan Tiantian, vice dean and professor at Tianjin University of Technology.
AI虚拟手语主播背后蕴含强大技术，确保其能够完成这项工作。手语语序不同于口语普通话。例如，听障人士把谓语放在句尾，而我们健全人把谓语放在句子中间，即现代汉语中的主语和宾语之间。因此，天津理工大学聋人工学院副院长袁甜甜教授表示，手语识别非常困难。
Yuan and her team conducted extensive research on action recognition. The arms have 18 points that need to be analyzed, a hand has 21 points and a face has more than 100. All of these points are challenging for the AI and algorithm to process.
袁教授和团队对手语动作识别进行了广泛研究。手臂有18个需要分析的点，一只手有21个分析点，一张脸有100多个分析点。所有这些点在人工智能和算法处理时都是挑战。
Compared with human language interpreters, the AI sign language anchor has some advantages. It can help with the continuous translation for long texts, and limit the amount of information lost. Also, statistics indicated that within special scenarios, the correct rate of sign language recognition could top 97 percent while the correct rate of sign language generation could be even higher.
与人类手语译员相比，AI手语主播有一些优势。AI主播可以助力长文本的连续翻译，还能限制信息丢失的数量。此外，统计数据表明，在特殊场景下，AI主播手语识别正确率最高可达97%，而手语生成正确率甚至更高。
After the Winter Paralympics, the AI sign language TV anchor will have no shortage of applications. The research team expects that it could help those hearing-impaired people in their shopping, transportation and other aspects of daily life.
北京冬残奥会结束后，AI手语主播还会应用在许多方面。科研团队希望AI主播能在购物、交通和日常生活的其他方面帮助听障人士。
China’s virtual human industry is ready for rapid growth, with applications in areas such as broadcasting, entertainment, retail, finance and education. It is also possible that everyone eventually will have their very own avatar.
中国的虚拟人产业已经为快速增长做好准备，比如在广播、娱乐、零售、金融和教育等等应用领域。有可能最终每个人都将拥有自己的虚拟化身。
