A WPT/NFC-Based Sensing Approach for Beverage Freshness Detection Using Supervised Machine Learning
A WPT/NFC-Based Sensing Approach for Beverage Freshness Detection Using Supervised Machine Learning
复制标题
DOI：
10.1109/jsen.2020.3013506
复制
发表时间：
2021-01
影响因子：
4.3
通讯作者：
Daniel Rodriguez;M. Saed;Changzhi Li
中科院分区：
文献类型：
--
作者：
Daniel Rodriguez;M. Saed;Changzhi Li
无线传感器的大规模部署是不断发展的物联网 (IoT) 行业的基本组成部分。因此，必须利用现有的硬件来实现新的传感功能，而无需添加或很少添加硬件。随着无线充电（WPT）和近场通信（NFC）成为智能手机的标准功能，本文研究了基于与智能手机兼容的WPT/NFC技术的饮料新鲜度传感。开发了饮料-线圈相互作用的电路模型，并分析和测试了不同性质（例如幅度、幅度、相位）的分类特征的性能。当使用 5 种不同类型的牛奶时，使用监督机器学习进行牛奶新鲜度分类时，准确率高达 96.7%；当仅使用 2% 脂肪牛奶进行分类时，准确率高达 100%。此外，使用奇异值分解 (SVD) 和箱线图分析将分类所需的射频带宽减少至 10 MHz，而不影响两种不同特征提取方法的分类精度。
The massive deployment of wireless sensors is a fundamental piece in the growing internet of things (IoT) industry. Therefore, it is imperative to use already existing hardware to realize new sensing functions with very few or no hardware added. As wireless power transfer (WPT) and near field communication (NFC) become standard features in smart phones, this article investigates beverage freshness sensing based on the WPT/NFC technology compatible with smart phones. A circuit model for the beverage-coil interaction was developed and the performance of features from different nature (e.g., magnitude, amplitude, phase) for classification was analyzed and tested. Accuracies up to 96.7% were achieved using supervised machine learning for milk freshness classification, when 5 different types of milk were used and up to 100% when just 2% fat milk was used for classification. Additionally, the radio frequency bandwidth needed for classification was reduced to 10 MHz using singular value decomposition (SVD) and boxplot analysis without affecting the classification accuracy for two different methods of feature extraction.