| Soybean paste is one of the four main traditional fermented soybean products in China,which is characterized with rich nutrition,delicious taste and unique flavor,and therefore is an important condiment in people's daily life.Flavor is the crucial quality index that influences consumers' acceptance of soybean paste and it is of great significance to achieve its scientific characterization for production regulation and quality management.At present,the evaluation and characterization of soybean paste flavor quality depends mainly on artificial sensory tests,traditional physicochemical and instrumental detection methods.Although the artificial sensory evaluation method is most preferred owing to its ability to reflect consumers' actual feelings,it has strong subjectivity and poor reproducibility.On the other hand,the traditional physicochemical and instrumental detection methods can objectively and accurately determine the composition of substances.However,they are tedious,time-consuming,labor-intensive,and unable to feedback flavor information.In this study,based on the systematic analysis of physicochemical components and sensory attributes that affect the comprehensive flavor of soybean paste,a variety of sensing technologies including electronic nose,electronic tongue,colorimeter and texture analyzer were used to comprehensively and quantitatively characterize the flavor quality such as odor,taste,color and texture of soybean paste.Furthermore,multi-sensor information of flavor quality of soybean paste was fused by the information fusion technology,so as to establish a comprehensive and quantitative characterization and evaluation method for the unique flavor of soybean paste base on the intelligent sensing technology.The main contents of this research are as follows:1.Research on artificial sensory evaluation and physicochemical quality detection of soybean paste(1)In accordance with the national standards for evaluating the sensory requirements of soybean paste in national standards,the artificial sensory scoring rules of soybean paste were formulated,and the soybean paste samples were artificially evaluated using the scoring method.The results were used to build the predictive models for the flavor properties of soybean paste based on intelligent sensing technology.(2)Gas chromatography-mass spectrometer(GC-MS)technology was used to determine the aroma components of soybean paste samples,and physicochemical detection method was used to determine total acid,reducing sugar,salt,and amino acid nitrogen of soybean paste.Then,the main aroma components and characteristic taste index of soybean paste was analyzed base on that.The obtained results were used to establish the quantitative prediction models for the physicochemical indexes related to unique flavor of soybean paste.2.Research on detection method of physicochemical indexes related to soybean paste flavor based on intelligent sensing technologiesElectronic nose and electronic tongue technology were used to collect the odor and taste sensing information of soybean paste and extract the eigenvalues respectively.Partial least square(PLS)and Support vector machine(SVM)methods were used to establish quantitative prediction models of the main aroma components and physicochemical indexes related to characteristic taste of soybean paste.Among them,the PLS prediction model of the main aroma components,total acid,reducing sugar,salt and amino acid nitrogen in the samples from the prediction set showed that the correlation coefficient(Rp)were 0.6660,0.9359,0.8431,0.4595 and 0.7972,and the root means square error of prediction set(RMSEP)were 8.6038,0.2330,2.9649,0.8565 and 0.0841,respectively;Likewise,the SVM prediction model yielded Rp of 0.8036,0.9494,0.9604,0.8969,0.9235,and RMSEP of 13.1902,0.1557,1.3881,0.2350 and 0.1615,respectively.The results showed that the prediction of the quality traits with the PLS model was not ideal,while the effect of the SVM model was better,having the correlation above 0.89(excluding the main aroma components).This showed that it is feasible to quantitatively predict the flavor components of soybean paste based on electronic nose and electronic tongue.3.Research on characterization method of sensory attribute of soybean paste flavor based on intelligent sensing technologiesFour intelligent sensing technologies,namely electronic nose,electronic tongue,colorimeter and texture analyzer were used to collect various flavor information of soybean paste and extract the eigenvalues as the model input to establish the quantitative prediction model for the sensory attribute of soybean paste.For the PLS prediction model,the Rp of aroma,taste,color,and texture scores of the samples from the prediction set were 0.8642,0.8454,0.5851,0.8661,and the RMSEP were 1.3702,1.7895,2.1015 and 2.0755,respectively;For the SVM prediction model,the corresponding Rp were 0.9002,0.9368,0.8074,0.8855,and the RMSEP were 1.1054,1.1382,1.0011 and 1.6082,respectively.The results showed that the correlation and stability of the SVM model are significantly better than the PLS model,the intelligent sensing technology can realize the quantitative prediction of the sensory attribute score of soybean paste,and can be used as an auxiliary means for sensory evaluation.4.Research on comprehensive characterization method of unique flavor of soybean paste based on intelligent sensing fusion technologyBased on the research of single flavor attributes,the qualitative classification and quantitative prediction model of comprehensive score for flavor of soybean paste were established based on the fusion of odor and taste sensing information and fusion of multi-sensor information,respectively.The results showed that the qualitative clustering results were consistent;and the quantitative prediction results for the PLS model based on the fusion of odor and taste sensing information yielded an Rp of 0.9150,with RMSEP being 3.9443.The Rp of SVM model was 0.9267,with RMSEP of 2.4002;For the PLS model based on the fusion of multi-sensor information,the Rp was 0.9339,with RMSEP of 3.5882,and the Rp of the SVM model was 0.9636,with RMSEP of 1.8506.The results showed that two fusion methods established can achieve the comprehensive characterization of the unique flavor of soybean paste,and the model based on the fusion of multi-sensor information has a higher accuracy and is relatively similar to human sensory perception.In this study,based on the purpose of comprehensively and quantitatively characterizing the flavor of soybean paste,its flavor components and sensory attributes were used as a control,and a variety of intelligent sensing technologies,namely electronic nose,electronic tongue,colorimeter and texture analyzer were used to evaluate the unique flavor of soybean paste of which good results were obtained.This study proved the feasibility of using intelligent sensing technology to characterize and evaluate the unique flavor of soybean paste.The evaluation models for the comprehensive flavor of soybean paste base on the odor-taste sensing information fusion and multi-sensor information fusion were compared,which provided a theoretical basis for the selection of soybean flavor evaluation model under different application scenarios and accuracy requirements.The research results can provide reference method and theoretical support for intelligent sensing technology to assist or even replace traditional flavor characterization methods. |
PDF Full Text Request