Validation of the rapid detection approach for enhancing the electronic nose systems performance, using different deep learning models and support vector machines

Real-time gas classification is an essential issue and challenge in applications such as food and beverage quality control, accident prevention in industrial environments, for instance. In recent years, the Deep Learning (DL) models have shown great potential to classify and forecast data in diverse problems, even in the electronic nose (E-Nose) field. In this work, a Support Vector Machine (SVM) algorithm and three different DL models were used to validate the rapid detection approach (based on processing an early portion of raw signals and a rising window protocol) over diverse measurement conditions. We performed a set of experiments with five different E-Nose databases, including fifteen datasets to be used with these algorithms. Based on the obtained results, we concluded that the proposed approach has a high potential and reduces the response time for making E-nose forecasts. Because in more than 60 % of the cases, it achieved reliable estimates using only the first 30 % or fewer of measurement data (counted after the gas injection starts). The findings suggest that the rapid detection approach generates reliable forecasting models using different classification methods. Moreover, SVM seems to achieve the best accuracy and better training time.