A Temporal Convolutional Network (TCN) is a class of deep neural network architecture that applies dilated, causal convolutions across the time dimension to capture sequential dependencies in time-series data, enabling efficient parallel processing of temporal patterns without the recurrence bottleneck of RNNs. In WiFi CSI sensing, TCNs are valued for their ability to extract multi-scale temporal features from dynamic signal streams such as gesture or activity sequences, making them well-suited for encoding compact temporal descriptors as in physics-aware frameworks like PULSE, or for processing time-varying CSI topology signals as in WiGNN. Key variants include stacked dilated convolutional blocks with residual connections, which extend the receptive field to capture both short-term fluctuations and longer-range temporal context relevant to recognizing complex human motions across diverse deployment conditions.
Source Papers
- PULSE: Physics-Aware Temporal Embedding Learning for Domain Adaptive Wireless Sensing ↗ — PULSE: Physics-Aware Temporal Embedding Learning for Domain
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired