An Intelligent Real-Time Odor Monitoring System Using a Pattern Extraction Algorithm

This study proposes an intelligent real-time odor monitoring system for monitoring odor problems, a new type of environmental problems, in real-time and for managing harmful substances. This system minimizes the measurement error of individual sensors by building a gas sensor array integrating 8 sensors, and applies a pattern recognition algorithm revised for accurate data analysis. The revised pattern recognition algorithm evaluates the similarity of patterns by combining ANN (Artificial Neural Networks) in testing the similarity and fitness of GA (Genetic Algorithm) in order to enhance the reliability of harmful gas pattern extraction. Moreover, the proposed system applies PCA (Principal Component Analysis) for the clustering of each gas pattern. What is more, this study designs a pattern matching algorithm and compares with data in the built DB in order to identify output malodorous substances. In addition, the outcomes are evaluated through the comparison of the ANN, GA and proposed ANN-GA algorithms. As the proposed system can enhance the reproducibility, reliability and selectivity of odor sensor output, it is expected to be applicable to diverse environmental problems including air pollution.