Abstract: In order to quickly understand the impact of objective sound environment in wetland park on people's subjective preference, sound pressure level data in the area and subjective score of soundscape preference are obtained through sound walking experiment. Taking the dominant sound of measurement points as the basis for the classification of soundscape types, 11 types of functions are selected to fit sound pressure level and soundscape preference score. Root mean square error (RMS), chisquare and Akichi information content criteria (AIC) are used to test the accuracy of the model and the fitting models with various kinds of soundscape preferences are superposed to construct the soundscape preference evaluation model (ESPWP model) of Wetland Park. At the same time, the model accuracy is tested by goodness of fit. The results show that the sound pressure level is negatively correlated with the scores of three kinds of soundscape preferences (P<0.01). The mean values of preference score of biological sound and earth sound are significantly higher than that of artificial sound, and the pressure level of artificial sound is significantly higher than that of earth sound and biological sound (P<0.05). In the fitting model of preference degree of each soundscape, the earth sound and biological sound are relatively inclined to the conic function model, and the artificial sound is inclined to the inverse function model, and the goodness of fit is 0.90, 0.81 and 0.87, respectively. The threshold values of sound pressure level of earth sound and biological sound are 43-45 dB (A), and the preference score of artificial sound decreases with the increase of sound pressure level. Due to the index values of model accuracy detection R²=0.73, the ESPWP model is a high-performance model, which can easily and quickly convert sound pressure level into subjective evaluation value of soundscape preference, to provide reference ideas and data support for the optimization and improvement of soundscape in wetland park.