Optimized Back-Propagation Combined with Radial Basic Neural Network for Improving Performance of the Electronic Nose: Case Study on the Fermentation Process of Tempeh

Tempeh is a popular traditional food from Indonesia, and its physical and chemical characteristics have been intensively investigated. Study on the fermentation process, however, still requires more attention. Here, the aroma profile during 10-day fermentation process was analyzed using an electronic nose (e-nose). The e-nose used in this study was based on eight kinds of the gas sensor (metal oxide semiconductor). A differential baseline manipulation and a maximum value as feature extraction were applied for the data processing. Meanwhile, the back-propagation neural network (BPNN) was used to classify the tempeh aroma profiles into four classes as described by principal component analysis (PCA). In this work, we optimized the BPNN and combined with the radial basis neural network (RBNN) to improve the performance of the e-nose. As a result, the combination between optimized BPNN and RBNN is able to recognize all data to the appropriate class as 100%.