Abstract: The aim of the study is to investigate the relationships between coffee aroma and banking conditions so as to provide theory reference for accurately selecting reasonable processing conditions for special fragrance of coffee and reducing the production cost of coffee diversification. The Hainan Arabica coffee was used as material, and we made use of the solid phase microextraction-gas chromatographic-mass spectrometric (SPME-GC-MS) and the electronic nose to detect the volatile compounds and characteristic aroma of coffee processed under several baking temperatures for 6 min. The results showed that a total of 43 kinds of volatile compounds were detected. When the baking temperature was lower than 120℃, volatile components of coffee mainly included alcohols, ethers and amine which slightly contribute to coffee aroma. With the temperature increasing, aromatic compounds were gradually generated from the pyrolysis of coffee endogenous compounds. Furthermore, pyrazine and pyrrole of coffee volatile compounds began to appear at 120℃. The contents of furan, aldehydes, pyrazine and pyrrole, such as 2-acetyl-5-methylfuran, 2,6-diethyl-pyrazine, 5-methyl-2-furancarboxaldehyde and 1-(2-furanylmethyl)-pyrrole, all peaked at 140℃. Parazole and imidazole, whose contents were less than 3%, only were detected at 160℃. At the same time, the content of volatile compounds changed irregularly because of the pyrolysis reaction of esters, phenols and ketones. After processing at different temperatures, the fragrance was composed of a large number of complex components which were able to lead to sensor inductive effect of electronic nose. Moreover, all of the response values were significantly different according to Tukey's multiple-range test at the P<0.05 level. However, part of the sensors of electronic nose could effectively analyze the variation of coffee characteristic flavor, such as the sensor T30/1, 70/2, PA/2, P30/2 and LY2/AA. Both the principal component analysis (PCA) and the discriminant factorial analysis (DFA) performed well when they were used to analyze coffee samples, with the first 2 principal components (PCs) explaining 100% and the first 2 factors explaining 100% of the variations of coffee aroma, which could effectively discriminate the coffee aroma with different roasting degrees. The hierarchical cluster analysis (HCA) successfully clustered the different types of coffee into 4 groups which included raw coffee, light roast coffee, medium roast coffee and dark roast coffee. The raw coffee could be divided into the first class which mainly clustered alcohol, esters and alkane. The second species, including 80℃ and 100℃ samples, was likely to cluster the first PC and the first discriminant factor, which was regarded as light roast coffee. A part of coffee may be classified as medium roast coffee at 120℃ and 140℃ owing to esters, phenols, aldehydes, amine, pyrazine and pyrrole. When processing at 160℃, volatile components of coffee mainly contained plenty of aromatic compounds such as heptyl ether, 4-ethyl-2-methoxyphenol, 5-methylfuran-2-carbaldehyde and 1-(furan-2-ylmethyl)-1H-pyrrole, in other words, this kind was categorized as dark roast coffee. It was observed that, the higher the baking temperature was, the more easily the samples were discriminated with different volatile compounds. Therefore, the relationships among the roasted coffee volatile compounds, the special fragrance and the roasting degree were determined by the SPME-GC-MS in combination with the electronic nose, which contributed to discriminate different roasting degrees of coffee. The research results can provide scientific basis and technical support for selecting the production process to roast particular aroma of coffee.