Study On Correlation Between Aroma Quality And Note Formula Of Blend Flavor

As the materials of which the flavor formula is usually made up are variety andcomplicated in mechanism of action, it’s difficult for qualitative or quantitative description inthe nonlinear relationship between aroma quality and note formula. In this paper three kindsof chemometrics ways were used to find the relationship between aroma quality and noteformula of blend flavor. The relationship was useful to guide and service for flavor blend andalso could achieve intelligent simulation of the flavor sensory evaluation and improve flavorresearch efficiency and level of technological innovation, on the other hand was significanceof environment by reducing material use and saving manpower and resources.First of all, modern flavor technology was used to prepare mushroom flavor by usingraw materials with reliable sources and stable quality permitted by GB2760-2011. Thecharacter note, sweet note, caramel note, soy sauce note, meet note and acid note werediscussed and then the diffuse fresh real and coordinate mushroom flavor was prepared. Thesensory evaluation technology was used to the aroma quality of mushroom flavor.Secondly, partical least squares (PLS), back propagation artificial neural network(BP-ANN) and support vector regression (SVR) were used to find the relationship betweenaroma quality and note formula of mushroom flavor. The result was that three kinds of modelscould all applicate in mushroom flavor as the sensory evaluation of mushroom was correlatedwith caluculated results, yet BP-ANN and SVR modles had got better results than PLS modlein leave-one-out cross validation (LOOCV) which meant the generalization ability. However,the BP-ANN modle had less accuracy of predicting than PLS and SVR modles. Datasuggested that use of SVR methods in relationship between aroma quality and note formula ofmushroom flavor was better than PLS and BP-ANN modles in modling, LOOCV andaccuracy of predicting.