A Recurrent Neural Network (RNN) is a class of deep learning architecture designed to process sequential and temporal data by maintaining hidden state information across time steps, allowing the network to capture dependencies in ordered input sequences. In WiFi/CSI sensing research, RNNs are particularly valuable for tasks such as people counting and activity recognition, where signal patterns evolve over time and temporal context is essential for accurate inference. Key variants commonly employed in this domain include Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRUs), both of which address the vanishing gradient problem inherent in standard RNNs and are better suited to learning long-range temporal dependencies in wireless signal data.

Source Papers

  • A low-cost automatic people-counting system at bus stops using Wi-Fi probe requests and deep learning — A low-cost automatic people-counting system at bus stops usi
  • Data Assimilation for Agent-Based Models — Data Assimilation for Agent-Based Models