Context-Aware Predictive Coding：面向 WiFi 感知的表示学习框架

WiFi 感知是一种 emerging 技术，利用无线信号 for 各种感知 application。然而，reliance on supervised learning、scarcity of labelled data、以及 incomprehensible channel state information (CSI) pose significant challenges。这些问题影响 deep learning 模型的 performance 和 generalization across different environments。Consequently，self-supervised learning (SSL) is emerging as a promising strategy to extract meaningful data representations with minimal reliance on labelled samples。在本文中，我们引入一种 novel SSL 框架，称为 Context-Aware Predictive Coding (CAPC)，有效地从 unlabelled data 中学习，并适应 diverse environments。CAPC 集成了 Contrastive Predictive Coding (CPC) 和 augmentation-based SSL 方法 Barlow Twins 的元素，promoting temporal 和 contextual consistency in data representations。这种 hybrid approach captures essential temporal information in CSI，这对于 tasks like human activity recognition (HAR) 至关重要，并且 ensures robustness against data distortions。此外，我们 提出了 unique augmentation，同时利用 uplink 和 downlink CSI 来 isolate free space propagation effects 并 minimize the impact of electronic distortions of the transceiver。我们的 evaluations demonstrate that CAPC not only outperforms other SSL methods and supervised approaches，但 also achieves superior generalization capabilities。Specifically，CAPC requires fewer labelled samples while significantly outperforming supervised learning and surpassing SSL baselines。Furthermore，our transfer learning studies on an unseen dataset with a different HAR task and environment showcase an accuracy improvement of 1.8 percent over other SSL baselines and 24.7 percent over supervised learning，emphasizing its exceptional cross-domain adaptability。