Daily Living Activity Recognition with Frequency-Shift WiFi Backscatter Tags

Area of Science:

• Ubiquitous Computing
• Wireless Sensor Networks
• Human-Computer Interaction

Background:

• Activity recognition is crucial for in-home services like elderly care.
• Radio-based methods (WiFi CSI, RFID, backscatter) offer privacy and low maintenance but face environmental and proximity challenges.
• Existing methods struggle with accuracy due to radio obstacles.

Purpose of the Study:

• To propose and evaluate a frequency-shift backscatter tag-based method for in-home activity recognition.
• To demonstrate the feasibility of this system in a realistic residential environment.
• To overcome limitations of existing radio-based sensing technologies.

Main Methods:

• Developed ultra-low power frequency-shift backscatter tags with antennas and switches.
• Implemented a sensing system using software-defined WiFi (SD-WiFi) access points and physical switches.
• Focused on detecting frequency shifts caused by tag movements for activity recognition.

Main Results:

• Frequency shifts detected within 2m with 72% accuracy under line-of-sight (LoS) conditions.
• Achieved 96.0% accuracy (F-score) in recognizing seven daily living activities with optimal receiver/transmitter setup.
• Demonstrated successful frequency shift detection without LoS at 3-5m by increasing overlaying packets.

Conclusions:

• Frequency-shift backscatter tags offer a robust and low-power solution for in-home activity recognition.
• The proposed system shows significant potential for enhancing elderly care and other in-home services.
• The method's accuracy and range can be improved, even in non-line-of-sight scenarios.