With the development of science and technology, the higher and higher surface accuracy of workpiece is required. The characterization of surface topography affect greatly on functional performance of workpiece. In recent years, many researchers have made a large progress on the study of surface topography and presented some related standards. The assessment of parameters of surface roughness, however, has not been researched intensively in the new generation of standards system. This paper proposes a novel roughness separation method based on B spline wavelet to analyze engineering surface. This method improves accuracy of roughness signal separation. The experimental results show that B spline wavelet can extracts accurately the roughness component and experiment results by the proposed method fit with the actual value of the standard sample plate.