The Research Of Taste Of Chewing Tobacco And The Application Of Rose Flavour In Chewing Tobacco

Under the situation that tobacco control is increasingly tough, chewing tobacco research and development have become the focus in the various tobacco giants, enhancing Chinese consumers' acceptance and taste of chewing tobacco and improving the falvour has become a new research target. we used the electronic touge to measure the characteristics of chewing tobacco and has established the mathematical model of chewing tobacco taste characteristics by using the SVR (support vector machine) and BP-ANN (artificial neural network). Using GC-MS combined with GC-O/AEDA at the same time to determine the characteristics of rose essential oil, and apply the recombinant rose flavour to chewing tobacco, and test the effects of the characteristics of chewing tobacco taste.11 kinds of different chewing tobacco samples were classified and tested using electronic tongue and got that: the electronic tongue has a good distinguish between different flavors,different production batches and the same flavor but different flavors(P<0.05) and show good stability and repeatability (RSD < 10). A mathematical model was established by testing 25 kinds ofchewing tobacco with different formulations and the result shows that: For sour taste model, the MRE values of SVR and BP-ANN were 4.64% and 1.71% respectively, and the the prediction error of two samples were 2.77% and 0.8% respectively, is greater than that of BP-ANN model, 2.22% and 1.6%, showing that the BP -ANN is more accurate. For sweet taste model, the MRE values of SVR and BP - ANN were 1.11% and 1.71 % respectively, and the the prediction error of two samples were 0.7% and 3.6% respectively, is lower than that of BP- ANN model, 2.6% and 4.7%, showing that the SVR is more accurate. For bitter taste model, the MRE values of SVR and BP - ANN were 4.64% and 1.71% respectively, and the the prediction error of two samples were 2.77% and 0.8% respectively, is lower than that of BP-ANN model, 2.22% and 1.6%, showing that the BP-ANN is more accurate.37 key volatiles (OAV>1 and FD?4) were detected through GC-MS combined with GC-O/AEDA and sensory evaluation. 3 volatiles (FD=4) including ?-pinene, respectively,heptanal, farnesol, 13volatiles (FD=16) including caryophyllene oxide, ?-ionone, phenylethyl alcohol, geranyl acetate, dipentene, benzaldehyde, nonyl aldehyde, alcohol, ether, rose methyl ketone of heptene,?-myrcene,?-pinene,valeraldehyde, 11 volatiles (FD=64) including lauric acid, capric acid and octanoic acid, 3-allyl-6-methoxy phenol, ?-geraniol, neryl acetate,?-terpineol, citronella acetate, citronollyl formate, camphene, hallow, geranyl acetone, 4 volatiles (FD=256) including benzyl benzoate, citronol, citral, and acetone, 6 volatiles(FD=1024) including a-terpinene, hexanol, citronellal, linalool, nerol,methyl eugenol.sensory evaluation and ENA analysis showed a significantdifference in the sweet ,sour and bitter taste of chewing tobacco (P < 0.05).