A 1D CNN (One-Dimensional Convolutional Neural Network) is a deep learning architecture that applies convolutional filters along a single temporal or sequential axis to automatically extract local patterns and features from time-series or vector-form CSI data, such as amplitude or phase sequences derived from channel frequency responses. In WiFi sensing, it matters because it offers a computationally efficient alternative to classical machine learning classifiers and more complex multi-dimensional deep learning models, enabling effective feature learning directly from raw or lightly preprocessed CSI signals without extensive manual feature engineering. Key variants in this domain include lightweight single-branch 1D CNNs used for tasks like crowd counting, as well as 1D convolutional components embedded within larger hybrid architectures that incorporate temporal embedding or domain adaptation modules, such as those operating on compact physics-aware descriptors rather than full CSI tensors.
Source Papers
- CSI crowd-counting: An experimental study using Machine Learning and Deep Learning Algorithms ↗ — CSI crowd-counting: An experimental study using Machine Lear
- PULSE: Physics-Aware Temporal Embedding Learning for Domain Adaptive Wireless Sensing ↗ — PULSE: Physics-Aware Temporal Embedding Learning for Domain