A Bidirectional Long Short-Term Memory (BiLSTM) network is a recurrent neural architecture that processes sequential CSI data in both forward and backward temporal directions, allowing the model to capture dependencies from past and future context simultaneously within a time series. In WiFi-based human sensing, this bidirectional processing is particularly valuable because human gestures and activities exhibit temporal patterns that are better characterized when the model can consider the full temporal context rather than relying solely on prior observations, as in a standard unidirectional LSTM. BiLSTM appears as a distinct variant within the broader family of recurrent models benchmarked for CSI sensing tasks, alongside vanilla RNNs and standard LSTMs, and is often paired with convolutional front-ends or attention mechanisms to further enhance its ability to extract discriminative spatiotemporal features from CSI amplitude and phase signals.
Source Papers
- Channel State Information from Pure Communication to Sense and Track Human Motion: A Survey ↗ — Channel State Information from Pure Communication to Sense a
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- Device-Free Wireless Sensing for Gesture Recognition Based on Complementary CSI Amplitude and Phase ↗ — Device-Free Wireless Sensing for Gesture Recognition Based o
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered
- WiFi-Based Human Sensing With Deep Learning: Recent Advances, Challenges, and Opportunities ↗ — WiFi-Based Human Sensing With Deep Learning: Recent Advances