Self-supervised learning is a training paradigm in which a model learns generalizable representations from unlabeled data by solving pretext tasks whose supervisory signals are derived automatically from the input data itself, without requiring manually annotated labels. In WiFi CSI sensing, this approach is particularly valuable because collecting large volumes of labeled CSI data is costly and environment-dependent, making it difficult to scale supervised methods across diverse real-world deployments in settings such as smart buildings. Key variants relevant to the field include contrastive learning methods, which train models to distinguish between augmented views of the same CSI sample, and masked autoencoding approaches, which reconstruct deliberately obscured portions of the CSI signal, both enabling robust feature learning that can be fine-tuned for downstream tasks like activity recognition or occupancy estimation with minimal labeled examples.
Source Papers
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered
- WiFi as Infrastructure: Valuation Impact of CSI Sensing on Smart Buildings and REIT Portfolios ↗ — WiFi as Infrastructure: Valuation Impact of CSI Sensing on S