Total synchronous fluorescence spectroscopy coupled with deep learning to rapidly identify the authenticity of sesame oil.

ABSTRACT

The quality of sesame oil (SO) has been paid more and more attention. In this study, total synchronous fluorescence (TSyF) spectroscopy and deep neural networks were utilized to identify counterfeit and adulterated sesame oils. Firstly, typical samples including pure SO, counterfeit sesame oil (CSO) and adulterated sesame oil (ASO) were characterized by TSyF spectra. Secondly, three data augmentation methods were selected to increase the number of spectral data and enhance the robustness of the identification model. Then, five deep network architectures, including Simple Recurrent Neural Network (Simple RNN), Long Short-Term Memory (LSTM) network, Gated Recurrent Unit (GRU) network, Bidirectional LSTM (BLSTM) network and LSTM fortified with Convolutional Neural Network (LSTMC), were designed to identify the CSO and trace the source with 100% accuracy. Finally, ASO samples were also 100% correctly identified by training these network architectures. These results supported the feasibility of the novel method.