In　the　scenario　of　the　coexistence　of　long　term　evolution　(LTE)　system　using　orthogonal　frequency　division　multiplexing　(OFDM)　and　narrowband　Internet-of-Things　(NB-IoT),　when　the　in-band　model　is　used,　the　narrowband　interference　caused　by　NB-IoT　will　seriously　degrade　the　performance　of　OFDM.　Meanwhile,　the　phase　noise　(PHN)　generated　by　the　local　oscillators　also　reduce　the　reliability　of　OFDM　transmission.　In　this　letter,　an　algorithm　of　jointly　eliminating　NB-IoT　interference　and　PHN　is　proposed　within　the　sparse　Bayesian　learning　framework　for　OFDM　system,　when　the　channel　noise　follows　non-Gaussian　distribution.　An　iterative　solution　is　derived　after　the　optimization　structure　is　constructed.　Simulation　results　demonstrate　that　the　gain　of　proposed　method　over　the　conventional　methods　is　approximately　1.5-3.5　dB　when　the　bit　error　rate　is　{10{-3}}.　©　2012　IEEE.