Modeling RFID Signal Reflection for Contact-free Activity Recognition

Wireless sensing techniques for tracking human activities have been vigorously developed in recent years. Yet current RFID based human activity recognition techniques need either direct contact to human body (e.g., attaching RFIDs to users) or specialized hardware (e.g., software defined radios, antenna array). How to wirelessly track human activities using commodity RFID systems without attaching tags to users (i.e., a contact-free scenario) still faces lots of technical challenges. In this paper, we quantify the correlation between RF phase values and human activities by modeling intrinsic characteristics of signal reflection in contact-free scenarios. Based on the signal reflection model, we introduce TACT that can recognize human activities using commodity RFIDs without attaching any RFID tags to users. TACT first reliably detects the presence of human activities and segments phase values. Then, candidate phase segments are classified according to their coarse-grained features (e.g., moving speed, moving distance, activity duration) as well as their fine-grained feature of phase waveform. We deploy and leverage multiple tags to increase the coverage and enhance the robustness of the system. We implement TACT with commodity RFID systems. We invite 12 participants to evaluate our system in various scenarios. The experiment results show that TACT can recognize eight types of human activities with 93.5% precision under different and challenging experiment settings.