An Odor Monitoring System Based on Differentiated Pattern Recognition Implemented a Semiconductor Gas Sensor Array

摘要

At room temperature or in the refrigerator, foods such as beef, fish and shellfish decompose very quickly and such decomposed foods are highly likely to cause food poisoning. In particularly, food kept in the refrigerator can be decomposed unknowingly to the user, and the decomposition can produce harmful gases such as volatile fatty acid, CO_2 and methane to cause new diseases. In this study, an odor monitoring system using a recognition module integrating 8 gas sensors and differentiated pattern recognition algorithm built by the research institute is developed in order to measure the degree of food decomposition. The problem of low selectivity, a common characteristic of sensor arrays, was solved through pattern recognition technology using the proposed Genetic Artificial Neural network (GANN) algorithm. The GANN algorithm evaluates the similarity of pattern using ANN in the similarity and fitness evaluation of GA in order to enhance the reliability and selectivity of pattern extraction from harmful gas. We analyzed the characteristics of a pattern for each input gas, and evaluated the results of matching with DB data built up for identifying output odorous substance. In the evaluation, the proposed GANN algorithm was compared with existing ANN and GA, and the proposed GANN showed the highest recognition rate with 97% matching. Through further studies, this idea can be applied not only to real-time monitoring of indoor fire, gas leaking, etc. but also to various prediction and forecasting systems of observed data in diverse areas including environment, medicine, survey, safety/security, chemistry and food.