摘要
现有双向手语自动翻译系统估算得到的阻抗参数过小,翻译的仿生手部滑动轨迹及节点过少的问题,导致翻译不准。设计基于机器学习的智能仿生双向手语自动翻译系统。设计智能仿生手主控控制器结构,并在传感器结构内,利用质量块及弹簧,组建陀螺仪压力结构。软件部分标定手语轨迹数据点,设定模拟数值控制关系,规整为双向的语义翻译过程,转化语言信息并输出编码向量后,利用编程语言实现自动翻译软件功能。随机选定十项基础手势后,采集手势翻译数据并整理为数据集,应用两种传统手语自动翻译系统与所设计的自动翻译系统进行实验,结果表明:所设计的手语自动翻译系统翻译处理的手部划定轨迹及节点最多,能够完整地翻译手语表示的语义。
The existing two-way sign language automatic translation system estimates the impedance parameters are too small,the translation of bionic hand sliding trajectory and nodes are too few,resulting in inaccurate translation.An intelligent bionic two-way sign language automatic translation system based on machine learning is designed.The main controller structure of intelligent bionic hand is designed,and in the sensor structure,the mass block and spring are used to build the gyroscope pressure structure.The software part demarcates the sign language trajectory data points,sets the simulation numerical control relationship,regularizes the two-way semantic translation process,transforms the language information and outputs the coding vector,and then realizes the automatic translation software function by using the programming language.After randomly selecting ten basic gestures,the data of gesture translation are collected and sorted into data sets.Two traditional sign language automatic translation systems and the designed automatic translation system are used for experiments.The results show that the designed sign language automatic translation system has the most hand delimitation tracks and nodes,and can translate the semantics of sign language completely.
作者
张雨曦
赵婉莉
ZHANG Yuxi;ZHAO Wanli(Public Course Teaching Department,Shaanxi Polytechnic Institute,Xianyang Shaanxi 712000,China;Foreign Language Department,Xianyang Normal University,Xianyang Shaanxi 712000,China)
出处
《自动化与仪器仪表》
2022年第1期176-180,共5页
Automation & Instrumentation
基金
陕西省“十三五规划”2020年度课题“POA视角下中华优秀传统文化融入英语教学的实践研究”(项目编号:SGH20Y1547)。
关键词
机器学习
智能仿生
双向手语自动翻译
阻抗参数
滑动轨迹
质量块
machine learning
intelligent bionic
automatic two-way sign language interpretation
impedance parameters
sliding track
mass block
