基于深度卷积神经网络的辅助虚拟训练动作检测系统设计  

Design of auxiliary virtual training action detection system based on deep convolution neural network

作　　者：於鹏[1] 张铭[1] 李海兵[1] YU Peng;ZHANG Ming;LI Haibing(Nanjing University of Technology,Nanjing 210009,China)

机构地区：[1]南京工业大学,江苏南京210009

出　　处：《现代电子技术》2022年第4期161-164,共4页Modern Electronics Technique

基　　金：江苏省自然科学基金(BK20181265);江苏省高校自然科学基金面上项目(18KJB520020)。

摘　　要：为了提高辅助虚拟训练动作检测系统的精度与有效性,文中提出一种基于深度卷积神经网络的辅助虚拟训练动作检测系统,该设计系统由逻辑层、技术层与应用层三部分组成。在系统硬件设计方面,设计虚拟显示模块、电源模块和检测数据采集模块,在提高系统稳定性的同时提高虚拟训练动作数据传输的精度。在系统软件设计部分,以深度卷积神经网络结构为基础,识别辅助虚拟训练动作,并采集辅助虚拟训练动作的图像,结合卷积层与全连接层对采集到的辅助虚拟训练动作图像进行处理,通过分类层输出辅助虚拟训练动作的识别结果。最后,根据训练动作的定位结果完成辅助虚拟训练动作的检测。实验结果表明,文中所设计的系统能够精准定位辅助虚拟训练动作,提高训练动作的检测精度,保障辅助训练的有效性。An auxiliary virtual training action detection system based on deep convolution neural network is proposed to improve the accuracy and effectiveness of the auxiliary virtual training action detection system.The designed system is composed of the logic layer,technology layer and application layer.In the design terms of the system hardware,the virtual display module,power supply module and detected data acquisition module are designed to improve the accuracy of data transmission of the virtual training action while improving the system stability.In the part of the system software design,based on the deep convolution neural network structure,the auxiliary virtual training action is recognized,and the image of the auxiliary virtual training action is collected.The collected auxiliary virtual training action images are processed in combination with the convolutional layer and the fully connected layer,and the recognition results of the auxiliary virtual training action are output by means of the classification layer.The detection of the auxiliary virtual training action is completed according to the positioning results of the training action.The experimental results show that the designed system can accurately locate the auxiliary virtual training actions,improve the detection accuracy of training actions,and ensure the effectiveness of auxiliary training.

关 键 词：辅助虚拟训练动作 深度卷积神经网络 系统设计 虚拟显示 检测数据采集 图像处理 训练动作检测 

分 类 号：TN911.23-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]