Novel Framework for Multi-Scale Occupancy Sensing for distributed monitoring in Internet-of-Things

Abstract Occupancy sensing is one of the integral parts of modern evolving security surveillance and monitoring system used over different types of infrastructure. With an aid of multiple form of occupancy sensors, the prime idea of occupancy sensing is to identify the presence or absence of occupants in specifically monitored area followed by transmitting back the sensing information either for storage or for prompting a set of commands from the connected control units. Review of existing schemes exhibits the presence of adoption of multiple methodologies over different variants of use-cases; however, they are quite case specific, uses expensive deployment process, and performs highly sophisticated operation. At present, there are no studies specifically reported of using multi-scale occupancy sensing suitable for large and distributed environment of Internet-of-Things (IoT). Therefore, the proposed study introduces a mechanism of novel multi-scale occupancy sensing considering a use case of smart university campus, although, it can be implemented over any form of different infrastructures too connected over IoT environment. The proposed scheme is implemented considering different types of cost-effective sensors, handheld devices and access points in order to identify the state of occupancy in large number of rooms present in the campus. The sensed data from distributed connected campus are aggregated over cloud server where they are subjected to suitable preprocessing to increase the data quality suitable for reliable prediction. Multiple set of potential learning-based schemes are integrated with proposed model to explore best fit model. This assessment scenario is not found reported in existing scheme to perform classification of states of occupancy. The study outcome shows Convolution Neural Network and Long Short-Term Memory to accomplish higher accuracy compared to other learning approach.