This　article　presented　a　novel　signal　processing　and　defect　recognition　method　in　MFL　inspection　system.　During　the　preprocessing　course,　time-frequency　analysis,　median　and　adaptive　filter,　and　interpolation　processing　are　adopted　to　preprocess　MFL　inspection　signal.　In　order　to　obtain　high　sensitivity　and　precision,　we　adopted　multi-sensor　data　fusion　technique　to　inspection　data.　A　wavelet　basis　function　(WBF)　neural　network　was　used　to　recognize　defect　parameters.　Through　constructing　a　knowledge-based　off-line　inspection　expert　system,　the　system　improved　its　defect　recognition　capability　greatly.　©　2010　IEEE.