手语数字人技术及智能应用研究  

Sign Language Avatar Technology and Its AI Application

作　　者：刘鸿宇[1] 付继林[1] 王珍珍[1] LIU Hong-yu;FU Ji-lin;WANG Zhen-zhen(School of Foreign Languages,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学外国语学院,秦皇岛066004

出　　处：《价值工程》2024年第2期98-101,共4页Value Engineering

基　　金：2018年国家社科基金一般项目《基于手语语料库的中国手语“体范畴”特点及其语言学价值研究》(18BYY081);2019-2020河北省高等教育教学改革研究与实践项目《基于OBE理念的手语多元素质人才培养研究》(2019GJJG065)。

摘　　要：对智谱AI、腾讯、华为等头部企业打造的手语数字人产品进行梳理,研究和探讨我国代表性手语数字人产品的主要技术路线和智能应用。手语数字人的制作流程包括形象设计、建模、动作捕捉、驱动、渲染。通过多模型、多算法等AI技术集成实现手语数字人应用场景落地。智谱AI手语数字人核心技术含seq2seq模型、生成对抗网络,创造性地将对比学习引入mBart模型等。当前我国手语数字人在新闻播报、教育、国际赛事和重要会议翻译等方面实现智能应用。未来我国手语数字人技术有望迎来新突破,智能应用也将拓展到智慧医疗、智慧交通等更多领域。Through examining the sign language avatars created by leading enterprises like Zhipu AI,Tencent,Baidu,and Huawei,this article probes into the sign language avatar technology and its AI application.Sign language avatar production procedure includes image design,modelling,motion capture,driving,rendering.Sign language avatar application is realized based on an integration of multiple models,multiple algorithms and many other AI technologies.Zhipu AI's core technology of sign language avatar consists of seq2seq model,Generative Adversarial Network,and creatively introducing comparative learning into mBart model.In the future,China's sign language avatar technology is expected to achieve new breakthroughs,and it AI application will be extended to more fields such as intelligent medical care and intelligent transport.

关 键 词：手语 手语数字人技术 智能应用 Seq2Seq模型 mBart模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]