Exploring Feasibility of Large-Area Photovoltaic Sheets as Indoor Interactive Surfaces.
Human activity recognition in indoor environments is useful for comfortable and efficient living and working in smart homes and buildings. Energy harvesting technologies such as the photovoltaics could offer advantages for low-cost installation, maintenance, portability and energy savings. In this work, we explore the feasibility of large-area indoor photovoltaic (PV) sheets for both energy harvesting and gesture recognition.
To study the feasibility of the PV sheet for gesture recognition, we crafted a prototype with a simpler structure for the sensor and electronics in comparison to related literature. The prototype consists of a large-area PV sheet (110mm width by 28mm height), a maximum power-point tracking (MPPT) module and a Lithium Polymer (LiPo) battery. Numerous hand gestures were performed above PV sheet by partially shadowing the PV cells. These shadows introduce anomalies in the photocurrent signal. Gesture pattern classification and recognition are performed on the photocurrent signals generated by the crafted prototype using machine and deep learning classifiers.
Experimental results show that with the proposed prototype it is possible to detect six distinct hand gestures with average overall accuracy of 86.1%. The best performing classifier, Random Forest, has achieved 97.2% overall accuracy results on testing data.
Proposal accepted for the CHI 2020 Workshop on self-powered sustainable interfaces and interactions (SelfSustainableCHI 2020).
SelfSustainableCHI 2020 Workshop Paper
SelfSustainableCHI 2020 Workshop Paper on GitHub
