摘要
目的:为了解决家用电子血压计自动检定系统因很难获取大量样本数据而导致品牌型号识别准确率不稳定的问题。方法:提出了一种基于生成对抗网络-卷积神经网络(GAN-CNN)的电子血压计品牌型号识别方法。在训练集中加入生成对抗网络(GAN)生成虚拟图像,利用卷积神经网络(CNN)提取特征并进行分类,利用数据增强、归一化等技术来提高模型准确性与稳健性。结果:实验表明,发现GAN-CNN模型采集响应成功并对电子血压计进行分类,而且准确度达到94.7%以上。相对于其他模型具有更高的识别准确度和更好的泛化能力。结论:本文所提出的基于GAN-CNN的电子血压计示数识别技术能够达到较高的准确率和鲁棒性,在智慧计量领域具有较好的应用前景。
Aims:This paper aims to solve the problem that the automatic verification system of household electronic sphygmomanometer is difficult to obtain a large number of sample data,which leads to the instability of brand and model recognition accuracy.Methods:A brand and model recognition method of electronic sphygmomanometer based on GAN-CNN was proposed.In the training set,virtual images generated by the generated antagonism network(GAN)were added;and the convolutional neural network(CNN)to extract features and classify them was used.Data enhancement,normalization and other technologies were used to improve the accuracy and robustness of the model.Results:The experiment showed that the GAN-CNN model was successful in collecting responses and classifying electronic sphygmomanometer;and the accuracy was more than 94.7%.Compared to other models,it had higher recognition accuracy and better generalization ability.Conclusions:The readout recognition technology of electronic sphygmomanometer based on GAN-CNN proposed in this paper can achieve high accuracy and robustness and has good application prospects in the field of intelligent measurement.
作者
江凯
杨凯
颜迪新
李子印
倪军
陈灿
JIANG Kai;YANG Kai;YAN Dixin;LI Ziyin;NI Jun;CHEN Can(School of Optical and Electronic Technology,China Jiliang University,Hangzhou 310018,China;Institute of Medical and Ionizing Radiation Metrology,Zhejiang Province Institute of Metrology,Hangzhou 310018,China)
出处
《中国计量大学学报》
2023年第3期429-436,共8页
Journal of China University of Metrology
基金
浙江省市场监督管理局NQI科研项目(No.20200105)。
关键词
电子血压计
生成对抗网络
卷积神经网络
GAN-CNN模型
信息识别
electronic sphygmomanometer
countermeasures network generation
convolution neural network
GAN-CNN
information identification
