基于残差网络的新型触觉传感器接触状态感知与识别  

Perception and Recognition of Contact State for a Novel Tactile Sensor Based on Residual Networks

作　　者：王菲露 吕珊娜 宋杨 李明坤 孙牛平 章英 章王勇 WANG Feilu;LÜShanna;SONG Yang;LI Mingkun;SUN Niuping;ZHANG Ying;ZHANG Wangyong(School of Electronic and Information Engineering,Anhui Jianzhu University,Hefei Anhui 230601,China)

机构地区：[1]安徽建筑大学电子与信息工程学院,安徽合肥230601

出　　处：《传感技术学报》2022年第10期1404-1411,共8页Chinese Journal of Sensors and Actuators

基　　金：安徽省自然科学基金面上项目(1908085MF204);安徽省住房城乡建设科技项目(2021-YF24,2022-YF168);安徽省质量工程项目(2020szsfkc0294,2021jyxm0269);安徽省高校优秀青年人才支持计划项目(gxyq2019056);安徽建筑大学质量工程项目(2020szkc03,2021jy70);2022安徽省新时代育人质量工程项目(研究生教育:44)。

摘　　要：基于触觉信息识别不同接触状态对于人机交互的实现具有重要意义。提出一种基于残差网络(Residual Networks)识别触觉传感器不同接触状态的方法。基于弹性体模板法制备了一款具有良好触觉感知能力的多孔石墨烯柔性触觉传感器。利用该触觉传感器采集了4种不同接触状态(拍打、重击、抚摸、推压)的时间序列数据,并对采集到的数据进行z-score标准化操作以及降采样处理。在此基础上,构建具有良好自适应性和泛化能力的残差网络模型对施加于该触觉传感器表面的4种接触状态进行分类识别,平均识别准确率为97.50%。实验结果表明:设计的多孔石墨烯柔性触觉传感器柔弹性好、灵敏度高、响应速度快,能够有效感知不同状态的接触力,残差网络模型可以高效地用于触觉传感器接触状态的分类识别。Recognizing different contact states based on tactile information is of great significance to realize human-computer interaction.A method based on the Residual Networks to identify different contact states of tactile sensors is proposed.Firstly,based on the elastomer template method,a porous graphene flexible tactile sensor with good tactile perception is prepared.After that,the tactile sensor is used to collect the time-series data of 4 different contact states(slap,thump,stroke,push).Then z-score standardization operation and dimensionality reduction processing are performed on the collected data.On this basis,a Residual Networks model with good adaptability and generalization ability is constructed to classify and recognize the four contact states applied to the surface of the tactile sensor.The recognition accuracy rate is 97.5%.The experimental results show that the porous graphene flexible tactile sensor designed has good flexibility,high sensitivity,fast response speed,and can effectively sense the contact force of different states.Residual Networks can be efficiently used to classify and recognize the contact state of the tactile sensor.

关 键 词：压阻式压力传感器 接触状态感知 残差网络 分类识别 

分 类 号：TP212.1[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]