Aiming　at　the　difficulty　of　radio　frequency　narrowband　interference　(NBI)　estimation　in　ultra-wideband　anti-jamming　communication,　considering　that　UWB　signal　is　extreme　narrow　pulse　while　NBI　is　continuous　wave　and　NBI　has　the　characteristic　of　sparse　representation　in　frequency　domain,　an　improved　orthogonal　matching　pursuit　(OMP)　algorithm　based　on　compressed　sensing　theory　is　proposed,　which　estimates　the　frequency　and　amplitude　of　NBI　through　setting　the　reconstruction　error　changing　rate　as　the　stopping　criterion　for　iteration　in　the　matching　process　under　a　sampling　rate　below　the　Nyquist　sampling　rate.　Theoretical　analysis　and　experiment　show　that　single-tone　or　multi-tone　NBIs　in　IR-UWB　system　can　be　estimated　effectively　using　sub-Nyquist　samples　of　received　radio　frequency　signal　with　fewer　iteration　number.