14 October 2024 in IEEE Transactions on Mobile Computing
Deep learning (DL) based Wi-Fi sensing has witnessed great development in recent years. Although decent results have been achieved in certain scenarios, Wi-Fi based activity recognition is still difficult to deploy in real smart homes due to the limited cross-environment adaptability, i.e. a well-trained Wi-Fi sensing neural network in one environment is hard to adapt to other environments. To address this challenge, we propose AdaWiFi, a DL-based Wi-Fi sensing framework that allows multiple Internet-of-Things (IoT) devices to collaborate and adapt to various environments effectively. The key innovation of AdaWiFi includes a collective sensing model architecture that utilizes complementary information between distinct devices and avoids the biased perception of individual sensors and an accompanying model adaptation technique that can transfer the sensing model to new environments with limited data. We evaluate our system on a public dataset and a custom dataset collected from three complex sensing environments. The results demonstrate that AdaWiFi is able to achieve significantly better sensing adaptation effectiveness (e.g. 30% higher accuracy with one-shot adaptation) as compared with state-of-the-art baselines.
3 September 2025 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Wireless sensing has emerged as a promising technology due to its inherently privacy-preserving and contactless characteristics, enabling a wide range of applications. Until now, a major challenge in translating research into real-world applications is achieving accurate data segmentation. This process involves identifying the start and end points of target activities within complex and extended time-series sensing data, forming the foundation for subsequent inference tasks. However, existing segmentation techniques in wireless sensing are often constrained to a specific signal type and exhibit limited performance in distinguishing fine-grained activities, particularly when faced with similar background signals. To address these issues, we present SegALL, a unified segmentation framework capable of processing diverse signals. By leveraging a lightweight, signal-independent deep learning architecture coupled with active learning, SegALL enables accurate segmentation for fine-grained activities, even under interference from activities with similar characteristics. Experimental evaluations on both synthetic and real-world datasets demonstrate that SegALL significantly improves segmentation reliability across various modalities, including IMU, Wi-Fi, and mmWave signals. For example, it achieves a segmentation accuracy of 91.12% for mmWave signals, outperforming previous segmentation solutions that achieved 70.50%. Furthermore, SegALL reduces manual labeling effort by 40.15% and maintains a lightweight computational cost, making it suitable for deployment on edge devices such as the Raspberry Pi.