A Machine Learning Approach to Indoor Occupancy Detection Using Non-Intrusive Environmental Sensor Data

Estimating Occupancy Levels in Enclosed Spaces Using Environmental Variables: A Fitness Gym and Living Room as Evaluation Scenarios

“…Of the multiple environmental variables that exist to determine the presence of human beings, temperature, humidity, and CO are among the most frequently used [ 1 , 5 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 18 , 20 , 21 , 22 ]. A reason for this is the large variety of sensors available on the market and their accessible cost when compared to more traditional solutions, such as weather stations.…”

“…Several systems are benefited by the occupancy information gathered in buildings, mainly the Heating, Ventilation, and Air Conditioners (HVAC) system. Similarly, real-time occupancy information can enable and benefit other building management services, such as security, emergency systems, fire systems, flow management, and automated energy management [ 5 , 8 , 9 , 10 ]. However, occupancy detection systems face several challenges, such as reliability, adaptability, and accuracy.…”

“…Some examples of these technologies are video cameras incorporating people counting methods and infrared illuminators, Radio Frequency Identification (RFID) sensors, optical tripwires, Wireless Fidelity (WiFi), and Bluetooth Low Energy (BLE). Although direct methods have proven to be a viable mechanism for measuring occupancy, they present serious disadvantages, such as expensive hardware, privacy issues, intrusiveness, complex processing, and installation feasibility [ 1 , 5 , 8 , 10 , 11 ]. On the other hand, indirect methods derive occupancy information by measuring the effects of human beings in an enclosed environment, these effects can be changes in CO concentration, temperature, humidity, luminosity, sound levels, among others.…”

“…IoT has increased drastically the number of devices connected to Internet and the computational power provided by them. Furthermore, one characteristic of the IoT devices is their significantly lower cost compared to equipment used in surveillance solutions or in existing commercial systems for occupancy detection [ 5 , 9 , 10 ]. This and other characteristics have led to the fast development of IoT, which in turn has driven Edge Computing (EC) [ 10 ].…”

“…For example, Parise et al [ 13 ] developed an open-source IoT architecture for occupancy detection based on temperature, humidity, and pressure as environmental parameters, and composed of an EC environment and a cloud-based infrastructure. Similarly, Zemouri et al [ 10 ] used EC devices and a cloud-based platform to detect occupancy in office buildings. In the same way, Adeogun et al [ 12 ] detected and estimated occupancy in an office building by implementing an IoT LoRa (low-power wide-area network protocol) monitoring system based on a combination of temperature, humidity, light, and CO (as environment parameters), nine different sensors, and more than 20 nodes.…”