Towards Activity Recognition Using Wi-Fi CSI from Backscatter Tags
Towards Activity Recognition Using Wi-Fi CSI from Backscatter Tags
复制标题
DOI：
10.1109/percomworkshops56833.2023.10150323
复制
发表时间：
2023-03
期刊：
影响因子：
--
通讯作者：
V. Erdélyi;Kazuki Miyao;Akira Uchiyama;T. Murakami
中科院分区：
文献类型：
--
作者：
V. Erdélyi;Kazuki Miyao;Akira Uchiyama;T. Murakami
最近，使用 Wi-Fi CSI 的活动识别因其较低的部署成本而受到广泛关注。然而，其性能取决于Wi-Fi设备的数量，在实际场景中可能会受到限制。为了应对这一挑战，我们提出了一种使用低成本反向散射标签的基于反向散射的 Wi-Fi CSI 提取方法。我们评估了使用反向散射 CSI 进行活动感测的可行性。我们的实验评估表明，可以在各个标签位置成功提取反向散射的 CSI。此外，我们对 3 个活动的实验表明，它们的 CSI 数据存在显着差异，并且可以使用机器学习对它们进行分类，准确率达到 94.3%，这表明使用反向散射 CSI 数据进行活动识别是可行的。
Recently, activity recognition using Wi-Fi CSI has received significant attention due to its low deployment cost. However, its performance depends on the number of Wi-Fi devices, which may be limited in practical scenarios. To address this challenge, we propose a backscatter-based Wi-Fi CSI extraction method using a low-cost backscatter tag. We evaluate the feasibility of using the backscattered CSI for activity sensing. Our experimental evaluation shows that the backscattered CSI can be successfully extracted in various tag locations. Additionally, our experiment with 3 activities shows that there are significant differences in their CSI data, and that they can be classified using machine learning with 94.3% accuracy, which suggests that activity recognition using backscattered CSI data is feasible.