Rupavatharam, Siddharth. Design and implementation of a sensitive ceiling based visible light sensor. Retrieved from https://doi.org/doi:10.7282/T3TB1BCF
DescriptionExisting photoelectric, pyroelectric infrared (PIR) and camera based solutions for applications such as motion detection, activity detection, occupancy detection and localization in indoor settings require people to carry a device on their person or dense deployment of transmitters and receivers or raise privacy concerns. Some if not all of these shortcomings can be overcome by using visible light. Leveraging visible light in an indoor setting enables reuse of pre-existing light sources as transmitters. The short wavelength (nanometers) of visible light, properties of reflection and creation of shadows ensure that light is contained to the room and makes eavesdropping hard. Any system which aims to make use of these properties of visible light first requires a sensor which can sense and convert visible light to a voltage. The work presented in this thesis presents all the steps that go into designing a sensitive ceiling based photoreceiver that is able to detect light level changes occurring on any floor including dark carpeted surfaces. The photoreceivers are designed to be placed along with the existing lighting fixtures in a room to make deployment as easy as fitting a light bulb. The photoreceiver is designed as a module that can be used as the first component of any visible light based system that wishes to detect faint changes in light. The efficacy of the photoreceiver in terms of sensitivity and accuracy is shown by testing it in a conference room under realistic use case scenarios and different floor types. Experiments are conducted by deploying the photoreceiver in a conference room on the ceiling at a height of 9 feet along with an LED lamp. Sensitivity of the photoreceiver is found by slowly dimming a single LED lamp until the conference room is completely dark. Photoreceiver analog to digital converter (ADC) readings corresponding to dimming light at the ceiling are noted. Accuracy of the photoreceiver is found by first making people walk through the sensing area and collecting raw readings. The raw readings are then processed using a detection algorithm. The photoreceiver is able to detect light level changes of the order of tens of lux occurring on the floor and detect motion of a person walking with an AUC-ROC score of 0.9739 while pointed at a bright floor surface and an AUC-ROC score of 0.9276 when pointed at a dark floor surface.