| Environments that feature "lighting systems that think" are becoming a reality through a fusion of advanced light sources, sensors, and integrated control systems. Pivotal to such system design is the concept of "smart rooms" that will be capable of automatically understanding different human usage patterns in a room (e.g., reading a book, working on a laptop, attending or presenting at a meeting, fall detection), and respond by providing the right light where and when needed, thereby serving the purpose of increased energy savings as well as improved health, comfort and productivity.;While "smart" rooms must understand where their occupants are and estimate what they are doing, in many environments (e.g., hospital rooms or restrooms), preserving the privacy of the occupants is a critical issue, so cameras cannot be used. On the other hand, commonly-used occupancy sensors such as passive infrared are often ineffective. In this thesis, we present a method for real-time person tracking and coarse pose estimation using a sparse array of low-cost, low-power, single-pixel time-of-flight (ToF) sensors mounted on the ceiling of the room. These single pixel sensors are relatively inexpensive compared to commercial ToF cameras, and unlike cameras, they preserve the privacy of the occupants, since the only information collected is a small set of distance measurements. The tracking algorithm includes higher level logic about how people move and interact in a room and makes estimates about the locations of people even in the absence of direct measurements. A maximum likelihood classifiier based on features extracted from the time series of ToF measurements is used for robust pose classification into sitting, standing and walking states. We evaluate and refine our algorithms in a real physical test bed called the Smart Space Test bed (SST) at the NSF Engineering Research Center for Lighting Enabled Systems and Applications (LESA) and also in a larger simulated lab environment built with the Unity game engine. |
