SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing

• Published in: Patterns (New York, N.Y.) Vol. 4; no. 3; p. 100703
• Main Authors: Yang, Jianfei, Chen, Xinyan, Zou, Han, Lu, Chris Xiaoxuan, Wang, Dazhuo, Sun, Sumei, Xie, Lihua
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 10.03.2023; Elsevier
• Subjects: activity recognition; benchmarking; channel state information; deep learning; gesture recognition; human sensing; transfer learning; ubiquitous computing; unsupervised learning; WiFi sensing; deep learning; transfer learning; WiFi sensing; gesture recognition; benchmarking; channel state information; ubiquitous computing; DSML 2: Proof-of-concept: Data science output has been formulated, implemented, and tested for one domain/problem; human sensing; activity recognition; unsupervised learning
• ISSN: 2666-3899; 2666-3899
• DOI: 10.1016/j.patter.2023.100703

Over the recent years, WiFi sensing has been rapidly developed for privacy-preserving, ubiquitous human-sensing applications, enabled by signal processing and deep-learning methods. However, a comprehensive public benchmark for deep learning in WiFi sensing, similar to that available for visual recognition, does not yet exist. In this article, we review recent progress in topics ranging from WiFi hardware platforms to sensing algorithms and propose a new library with a comprehensive benchmark, SenseFi. On this basis, we evaluate various deep-learning models in terms of distinct sensing tasks, WiFi platforms, recognition accuracy, model size, computational complexity, and feature transferability. Extensive experiments are performed whose results provide valuable insights into model design, learning strategy, and training techniques for real-world applications. In summary, SenseFi is a comprehensive benchmark with an open-source library for deep learning in WiFi sensing research that offers researchers a convenient tool to validate learning-based WiFi-sensing methods on multiple datasets and platforms. [Display omitted] •SenseFi offers a model zoo and a comprehensive benchmark for WiFi sensing•Pre-training models on large datasets can be generalized to downstream WiFi-sensing tasks•Shallow models outperform very-deep models across different environments•Processed datasets for different WiFi-sensing platforms are available to use WiFi is extensively used in wireless communication to connect devices within a network. Along with the development of machine-learning algorithms and the widespread use of Internet of Things (IoT) products, applications of WiFi have recently expanded from communication to sensing. The presence or motion of humans or any objects within the wireless environment can be interpreted from signal propagation patterns. Compared with traditional video forms of sensing, WiFi sensing has the benefits of privacy protection, non-line-of-sight (NLOS) detection, and broad coverage. A comprehensive WiFi-sensing framework called SenseFi is proposed in this article that integrates hardware platforms, learning algorithms, and datasets applied for different WiFi-sensing tasks. We hope that SenseFi will contribute to future algorithm design and evaluation for real-world applications. WiFi sensing is a method to detect the presence or motion of humans or objects within a wireless network by analyzing signal propagation changes utilizing learning algorithms. SenseFi, presented in this article, is a comprehensive framework for WiFi-sensing researchers. It integrates current learning algorithms, hardware platforms, and datasets for different WiFi-sensing tasks, which could support future research on designing WiFi-sensing models.