A characterizable and quantitative analysis for rice sensory property using multifrequency large amplitude pulse electronic tongue

Abstract

A characterizable and quantitative analysis for rice sensory property including the fragrance, appearance, palatability and texture was developed using multifrequency large amplitude pulse (MLAP) electronic tongue. Signal correction and Fast Fourier Transform were employed. It was resulted that the correction indexes of the original signals of 1 Hz, 10 Hz and 100 Hz were 10%, 7.5% and 10% respectively. The 460 × 6 × 3 corrected signal data was obtained, and the first 3 amplitudes and density amplitudes were extracted as the effective characteristic values. Three interactive induction units were constructed and interactive induction methods were established. The sample induction diagram and sensory property induction diagram were drawn. The prediction error of the model by the interactive induction method was within 5.0%. The correlation coefficients of the fragrance, appearance, palatability and texture were 0.9571, 0.9021, 0.9721 and 0.9607, respectively, while the total score was the largest (r2 = 0.9752). Compared to the neural network, the higher correlation coefficient of the sensory property model by the interactive induction method displayed a better ability to quantitatively predict the sensory property of rice.