Objectives: This paper focused on evaluating and exploratory analyzing applicability of multifrequency amplitude pulse electronic tongue in determination of sodium metabisulfite in shrimp. Methods: The extract of shrimp (Penaeus vannamei) with 20% alcohol aqueous solutions was subjected to signal acquisition of electronic tongue. Once electrode combinations were determined， extreme points， inflection points and fusion of extreme points with inflection points in resulting signals were used as feature to prepare corresponding dataset. Data were processed by different chemometrics approaches including principal components analysis(PCA)， partial least squares(PLS)， principal components regression (PCR) and artificial neural network(ANN). Their performances were evaluated by predicted values vs measured values plots， root mean square error and correlation coefficient. Results: The best electrode combination was determined as Pt (1 Hz)， Au (100 Hz)， Pd (10 Hz) and Ag (100 Hz). Inflection points exhibited a good performance in both qualitative classification and quantitative forecasting， while ANN performed best for regression models among three algorithms. ANN model using inflection points data possessed a RMSE(TE) of 0.106 and R2(TE) of 0.996 in test dataset， suggesting its effectiveness in predicting concentrations of sodium metabisulfite with advantage in both data size and predicting performance. Conclusions: It is feasible to determine concentrations of sodium metabisulfite by multifrequency large amplitude pulse electronic tongue. This study provided an alternative way for high throughout determining illegal addition of sodium metabisulfite.