With the development of ubiquitous computing and IoTs technology, intelligent sensing is becoming one of the key technologies of smart IoTs. Our institute is conducting research in novel intelligent sensing theories and technologies. In recent years, we focus on contactless sensing of human body using ubiquitous wireless signals such as Wi-Fi and 4G/5G. Regarding the fact that there lacks a general theory in the wireless sensing area, we have introduced the Fresnel zone model, originally proposed for light wave propagation, into the wireless sensing area. We have revealed the theoretical foundation, working mechanism, and sensing limit of contactless sensing with wireless signals like Wi-Fi and 4G/5G. We have also found that the signal patterns caused by both small and large human activities are related to the contextual factors such as position, orientation, and action range. The theory and the findings are now widely recognized and adopted by worldwide researchers.
Based on the Fresnel sensing model and theory, our team proposed a series of key technologies for Wi-Fi based contactless sensing. In terms of decimeter-level coarse-grained activity recognition, we have proposed real-time and continuous activity splitting, behavioral feature extraction, and moving direction detection technologies. We have also developed the first real-time indoor fall detection system and human moving direction detection system. In terms of millimeter-level fine-grained activity recognition, we have uncovered the sensing mechanism of micro-activity, proposed the key methodology for reliable detection of micro-activity, and addressed the fundamental question about when micro-activity is detectable. Meanwhile, for the first time, we have studied the detectability of millimeter-level activities (e.g., respiration) using WiFi signals and explained why the received signal may have wavelength fluctuations reflecting object movements. Based on these findings, we have proposed and implemented a reliable wireless sensing method for respiration detection. Representative works have been published in flagship journals including IEEE Computer, IEEE Communications Magazine, IEEE Transactions on Mobile Computing, and top conferences including ACM UbiComp.
Based on the new theory and technologies, our team has developed and deployed a series of systems for human behavior recognition using commercial Wi-Fi devices, including fall detection, respiration detection, sleep monitoring, intrusion detection, indoor location/tracking, gesture recognition, fitness recognition, and elder daily activity recognition. At the same time, we have applied for more than 20 domestic or international patents, and taken the lead to start wireless sensing application of smart nursing home and real-time human status monitoring in the world.