摘要
为了提高肌电机械手手势识别的精度,首先将肌电信号信息转换为图片,实现更直观、更明显的肌电信息展示,然后使用迁移学习微调Res Net50模型,构建基于肌电信号灰度图的肌电机械手手势识别模型,并使用该模型对三种手势进行识别实验。结果显示,ResNet50模型对基于肌电灰度图的手势识别率高达93%,有效减少了过拟合现象并提高了识别精度。
This study aims to convert electromyography(EMG)signals into two-dimensional grayscale images and use grayscale images and deep learning to improve the accuracy of EMG-controlled robotic hand gesture recognition.Firstly,the EMG signal information is transformed into images to achieve a more intuitive and clear display of EMG information.Then,transfer learning is used to fine-tune the ResNet50 model,constructing an EMG-controlled robotic hand gesture recognition model based on EMG grayscale images.This model is used to recognize three types of hand gestures,and the experimental results show that the ResNet50 model achieves a recognition rate of 93%for EMG grayscale-based gesture recognition.This effectively reduces overfitting and improves recognition accuracy.
作者
罗肖媛
黄文静
曾学武
段小刚
郑哲文
王子怡
LUO Xiao-yuan;HUANG Wen-jing;ZENG Xue-wu;DUAN Xiao-gang;ZHENG Zhe-wen;WANG Zi-yi(School of Materials Science and Engineering,Central South University of Forestry&Technology,Changsha Hunan 410004;Hetian(Hunan)International Engineering Management Co.,Ltd,Changsha Hunan 410006;Central South Intelligence Hunan Industry 4.0 Innovation Center,Changsha Hunan 410000)
出处
《长沙航空职业技术学院学报》
2023年第4期33-38,共6页
Journal of Changsha Aeronautical Vocational and Technical College
基金
湖南省自然科学基金面上项目“基于概率模糊理论的肌电假肢手抓握模式与力解码研究”(编号:2022JJ31015)阶段性研究成果。
关键词
肌电信号
灰度图像
卷积神经网络
手势肌电识别
electromyography(EMG)signals
grayscale images
convolutional neural networks
EMG gesture recognition