The University of Patras has proceeded to another great achievement within the context of TwinAIR project!
The conference proceedings from the 16th International Conference on Ambient Systems, Networks, and Technologies (ANT) have been published by the ELSEVIER journal.
The publication titled “A Novel Approach to Activity Recognition Based on Indoor Air Pollutant Data”, authored by M. Mentou, G. Protopsaltis, C. Mountzouris, and J. Gialelis and its respective DOI is the following: https://doi.org/10.1016/j.procs.2025.03.078
The abstract and the keywords of the publication can be found below:
Abstract
Indoor activities, such as cooking and cleaning, are significant sources of air pollutants, including volatile organic compounds (VOCs), carbon dioxide (CO₂), nitrogen oxides (NOx), and particulate matter (PM), all of which pose long-term health risks. This study investigates the relationship between pollutant emissions and specific activities by analyzing their temporal concentration patterns. Controlled experiments were conducted to measure pollutant levels before, during, and after various cooking and cleaning tasks.The collected data were formulated as a classification problem, and machine learning algorithms such as Random Forest, GRU, LSTM, and CNN were applied to identify activity patterns. Due to the sequential nature of the data, GRU and LSTM models outperformed the others and were further optimized through hyperparameter tuning. The findings reveal that air pollutant measurements can serve as reliable indicators of specific activities, introducing a new pathway for activity recognition in indoor environments.
Keywords
Air Pollutants, IAQ Wearable devices, Machine Learning.
You can reach the full article following this link: https://www.sciencedirect.com/science/article/pii/S1877050925008142