The loss of life and property caused by typhoons in the southeastern coastal regions of China is mainly attributed to the destruction of low-rise buildings. Therefore， analyzing the wind pressure characteristics on the roof surface under strong winds is of practical value in wind engineering. Based on the fractal theory of physics， this study conducts the fractal analysis of wind pressure pulse signals. The wind pressure and referenced wind velocity data were collected during Typhoon “Muifa” in 2011 from the Pudong experimental base of Tongji University. In the fractal analysis， the box-counting method was used to estimate the fractal dimension value for each wind pressure sample. In the result section， an example is given to demonstrate the positive correlation between fractal dimension values and the fluctuation strength of wind pressure. Then， the fractal characteristics of wind pressures on the entire roof region and the influence of roof pitches and inflow directions on the fractal dimension were investigated. Firstly， the results indicate that the wind pressures under typhoon climate are anti-persistent time series， and the mean fractal dimension for wind pressures on the entire roof region is 1.700. Secondly， the roof pitch and inflow direction affect the fractal characteristics significantly. Thirdly， the peak region of the fractal dimension appears in the area far away from the incoming flow and close to the mountain wall， while local low values occur in the central area of the roof ridge. An analysis is carried out on the correlation between the fractal characteristics and the Non-Gaussian features. The negative correlation between fractal dimension and kurtosis coefficient is strong， with a negative correlation coefficient of -0.509 under case A-00.