You are here

Behavior Recognition based on Wi-Fi Channel State Information (CSI)

Feature Topic

CALL FOR PAPERS

Human behavior recognition is the core technology that enables a wide variety of human-machine systems and applications, e.g., health care, smart homes, and fitness tracking. Traditional approaches mainly use cameras, radars, or wearable sensors. However, all these approaches have certain disadvantages. For example, camera-based approaches have the limitations of requiring line of sight with enough lighting and potentially breaching human privacy. Low cost radar-based solutions have limited operation range of just tens of centimeters. Wearable sensor-based approaches are inconvenient sometimes because of the sensors that users have to wear. Recently, Wi-Fi CSI-based human behavior recognition approaches are attracting increasing attention. The rationale is that different human behaviors introduce different multi-path distortions in Wi-Fi CSI. Compared with traditional approaches, the key advantages of Wi-Fi CSI-based approaches are that they do not require lighting, provide better coverage as they can operate through walls, preserve user privacy, and do not require users to carry any devices as they rely on the Wi-Fi signals reflected by humans. As a result, the recognition of quite a number of behaviors that are difficult based on traditional approaches have now become possible, e.g., fine-grained movements (e.g., gesture and lip language), keystrokes, drawings, gait patterns, vital signals (e.g., breathing rate and heart rate), etc. However, Wi-Fi CSI-based behavior recognition still faces a number of challenges: How to build the CSI-behavior model and algorithms that are robust for different humans? How to overcome the impact of noise and ensure the performance of CSI-enabled systems? How to simultaneously recognize the behavior of multiple users? How the CSI-enabled system can adapt and evolve according to the environment change?

This FT provides the opportunity for researchers and product developers to review and discuss the state-of-the-art and trends of Wi-Fi CSI-based behavior recognition techniques and systems.

In the light of the above, the main goals of this FT are threefold:

  • to promote unparalleled / first-time approaches and techniques in signal processing, feature extraction, data mining and model construction for behavior recognition based on Wi-Fi CSI;
  • to identify open issues which remain a challenge towards the convergence of computation theories and technologies for behavior recognition based on Wi-Fi CSI;
  • to exploit novel application areas and demonstrate the benefits of Wi-Fi CSI in contrast with more traditional sensing approaches.

Topics may include (but are not limited to):

  • Behavior Recognition Model/Theory based on Wi-Fi CSI
  • Behavior Recognition Algorithms based on Wi-Fi CSI
  • Wi-Fi CSI Signal Processing for Behavior Recognition
  • Wi-Fi CSI Data Mining for Behavior Recognition
  • Novel Behavior Recognition Applications/Systems Supported by Wi-Fi CSI
  • Evaluation Metrics and Empirical Studies of Wi-Fi CSI enabled Systems
  • Quality-enhanced and adaptive sensing models with Wi-Fi CSI

GUEST EDITORS

Bin Guo (Corresponding Guest Editor)                  
Northwestern Polytechnical University, China           
guobin.keio@gmail.com   

Jennifer Chen
Stevens Institute of Technology, USA
yingying.chen@stevens.edu

Nic Lane
Bell Labs & University College London, UK
niclane@acm.org

Yunxin Liu                                                        
Microsoft Research Asia, China
yunxin.liu@microsoft.com

Zhiwen Yu
Northwestern Polytechnical University, China
zhiweny@gmail.com