Prediction of diesel particulate filtration is typically made by virtue of modeling of particulate matter(PM) collection. The model is closed with filtration parameters reflecting all small scale phenomena associated with PM trapping, and these parameters are to be traced back by inversely analyzing large-scale empirical data-the pressure drop histories. Included are soot cake permeability, soot cake density, soot density in the porous filter wall, and percolation constant. In the present study, a series of single channel DPF experiment is conducted, pressure histories are inversely analyzed, and the essential filtration parameters are deducted by DPF filtration model formulated with non-linear description of soot cake regression. Sensitivity analyses of model parameters are also made. Results showed that filtration transients are significantly altered by the extent of percolation constant, and the soot density in the porous filter wall is controlling the filtration qualities in deep-bed filtration regime. In addition, effect of soot particle size on filtration quality is distinct in a period of soot cake regime.