Characterization of selected Chinese soybean paste based on flavor profiles using HS-SPME-GC/MS, E-nose and E-tongue combined with chemometrics

•Characterizing commercial soybean paste flavor by intelligent sensory technology.•Prediction of esters in soybean pastes by gas sensors signals was proved feasible.•The differentiation performance of LDA were improved by feature-level data fusion.•The SVR models achieved a better predictive performance than PLSR models. Headspace solid-phase microextraction gas chromatography–mass spectrometry (HS-SPME-GC/MS) with electronic nose (E-nose) and electronic tongue (E-tongue) was applied for flavor characterization of traditional Chinese fermented soybean paste. Considering geographical distribution and market representation, twelve kinds of samples were selected to investigate the feasibility. A total of 57 volatile organic compounds (VOCs) were identified, of which 8 volatiles were found in all samples. Linear discrimination analysis (LDA) of fusion data exhibited a high discriminant accuracy of 97.22%. Compared with partial least squares regression (PLSR), support vector machine regression (SVR) analysis exhibited a more satisfying performance on predicting the content of esters, total acids, reducing sugar, salinity and amino acid nitrogen, of which correlation coefficients for prediction (Rp) were about 0.803, 0.949, 0.960, 0.896, 0.923 respectively. This study suggests that intelligent sensing technologies combined with chemometrics can be a promising tool for flavor characterization of fermented soybean paste or other food matrixes.