Estimating link travel time in a reasonable fashion based on multi-source data is becoming a major challenge for intelligent transportation system (ITS).In this study, five crucial parameters from multi-source data (i.e.data from fixed sensors and probe vehicles) were proposed by analyzing their impacts on link travel time.As a typical multisource data fusion (MDF) method, a three-layer Back-propagation Neural Network (BPNN) model was developed to estimate link travel time using different combinations of the proposed parameters as the models input vectors.To validate the BPNN model, estimated link travel time was compared to simulated link travel time obtained by Vissim-based experiments.Results showed that the developed model has good performance in time estimation, and reasonable input parameters of the model could improve estimation accuracy and constancy.