基于人工智能的手语翻译系统实现  

Implementation of Sign Language Translation System Based on Artificial Intelligence

作　　者：刘继兴 周昕[1] 张帅峰 曾令辉 段珍灵 沈顺权 LIU Jixing;ZHOU Xin;ZHANG Shuaifeng

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《科技创新与应用》2022年第23期41-43,48,共4页Technology Innovation and Application

基　　金：黑龙江省大学生创新创业项目(202110214035)。

摘　　要：听力障碍者通过手语与人进行交流,但手语并不被大众所掌握,为此提出一种手语翻译系统为听障人士和非听障人士的沟通提供便利。该手语翻译系统是一个基于人体姿态研究与手部识别的图像识别系统。基于帧间差分算法在连续手语视频帧中提取出关键帧,将帧数据上传到云端,然后在云端使用YOLOv5和Open Pose人体姿态识别开源库提取手语图像坐标信息,将坐标信息作为手语识别算法输入,得到预测结果,并以文本形式表现出来。People with hearing impairment communicate with able-bodied people through sign language, but sign language is not mastered by the public. For this reason, a sign language translation system is proposed to facilitate the communication between the hearing impaired and the non-hearing impaired. This sign language translation system is an image recognition system based on human posture research and hand recognition, key frames are extracted from continuous sign language video frames based on inter-frame difference algorithm, upload the frame data to the cloud, and then use YOLOv5 and OpenPose human posture recognition open source database to extract sign language image coordinate information, The coordinate information is used as the input of the sign language recognition algorithm, and the prediction results are obtained and expressed in the form of text.

关 键 词：手语翻译 人工智能 图像识别 云端服务 系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]