Based on the daily maximum surface air temperature data of 35 stations during 1961-2008 in the Three Gorges area, the climatic characteristics of high temperature are analyzed. Using the empirical mode decomposition, we have discussed the mean annual high temperature days as well as the vacillation mode structure characteristics of the mean daily highest temperature during annual high temperature days in the Three Gorges area in recent 48 years, and we constructed a new prediction model by using function extension of mean value generated by generating function model as well as BP nerve net. The prediction model made the projection of the mean annual high temperature days and the mean daily highest temperature. The results show that the mean annual high temperature days were 24.07 and the mean daily highest temperature during annual high temperature days was 36.69℃ in the past 48 years. Meanwhile, the mean annual high temperature days were negative anomaly in the 1980s and 1990s, but positive anomaly in the other periods of time, oscillating with a cycle of about 3-5 and 16 years. The mean annual high temperature days were more in the northern part, but less in the southeast part of the Three Gorges area. The mean daily highest temperature during annual high temperature days was lower in southeast part, but higher in northeast part of the Three Gorges area. By comparing the prediction value of the model with real value, it showed that the model not only could fit the history in the Three Gorges area, but also could predict the trend of the future 5-10 years in the Three Gorges area. However, the model should improve the prediction ability of the extreme climate events occurring occasionally.