一种基于YOLOv5的中国传统手语拼音数据库模型  

A Chinese Traditional Sign Language Pinyin Database Model Based on YOLOv5

作　　者：张笑尘 雷安旭 肖肃 施传威 王中天 王勤林 ZHANG Xiaochen;LEI Anxu;XIAO Su;SHI Chuanwei;WANG Zhongtian;WANG Qinlin(School of Artificial Intelligence,Jianghan University,Wuhan 430056,China)

机构地区：[1]江汉大学人工智能学院,湖北武汉430056

出　　处：《电视技术》2023年第4期38-42,共5页Video Engineering

摘　　要：针对中国手语在人工智能领域开发率低、数据库较少等问题,设计一款以中国聋哑人初学的手语拼音作为数据源的数据库,使用labelimg软件对图像进行特征标注,使用YOLOv5目标检测方法进行模型训练。设计的模型具有较高的精确率、召回率和检测准确率,图像识别率约为93%,可以完成与聋哑人无障碍交流,并且制作成本较低、模型易移植,适合在社会中进行推广普及。In response to the low development rate of Chinese sign language in the field of artificial intelligence and the lack of databases,this article designs a database using Chinese sign language pinyin as the data source,which is initially learned by deaf and mute people in China.Labelimg software is used for feature annotation of images,and YOLOv5 object detection method is used for model training.The model designed in this article has high accuracy,recall,and detection accuracy,with an image recognition rate of about 93%.It can achieve accessible communication with deaf and mute people,and has low production costs.The model is easy to transplant,making it suitable for promotion and popularization in society.

关 键 词：深度学习 机器学习 视觉识别 YOLOv5 手语数据库 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]