Internet of Things Enabled Smart E-Nose System for Pollutants Hazard Detection and Real-Time Monitoring in Indoor Mosquito Repellents

Mosquito-repellent is one of the most utilized products for dispelling mosquitoes that produce a significant quantity of gaseous, and particulate pollutants and are hazardous to human health. An indoor chamber experiment is conducted to identify the primary pollutants from mosquito repellents. This paper proposes an internet-of-things (IoT) based E-nose for volatile-organic compounds (VOCs)/gases/odors detection and real-time monitoring released from different mosquito repellents used in indoor environments. E-noses generate unique signature patterns for various VOCs/gases/odors. A gas sensor node interface is developed using eight cross-selective tin-oxide (MOX) based gas sensor arrays with a low-powered microcontroller to capture real-time signature patterns of VOCs/gases/odors released by different repellents when they are put in use. This microcontroller sends the gas sensor responses to the Amazon web services (AWS) cloud platform, from where data is ported to a remote data processing station (RDPS) for further analysis for its use in real-time. In this experiment, seven different types of mosquito repellents, viz mosquito coil, fast card, herbal cones, herbal spray, mosquito spray, hit spray and all-out liquid have been used. A dataset consisting of 1200 samples was captured. These sensor array responses are pre-processed using the two-stage analysis space transformation method, i.e., standardised linear discriminant analysis (SLDA) is used in the first stage. Subsequently, the second processing stage used Adaboost, random forest and recursive discriminant analysis (RDA) classifiers. The proposed e-nose was tested using 40 unknown VOCs/gases/odors samples that were not used during the training and validation. Experimentally, the RDA classifier trained in the SLDA transformed dataset could classify all 40 test samples with 100% accuracy. The lowest mean squared error achieved was 2.40 × 10–7. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.