A　fast　channel　prediction　scheme　is　proposed　for　Ricean　fading　scenarios.　By　employing　an　echo　state　network　(ESN),　the　scheme　is　able　to　obtain　smaller　prediction　error　than　previous　designs.　Simulation　results　show　that　the　ESN　prediction　method　has　lower　normalized　mean　squared　error　than　the　traditional　autoregressive,　discrete　wavelet　transform,　and　support　vector　machine　prediction　approaches.　The　symbol　error　rate　gap　between　the　perfect　and　predicted　channel　state　information　is　small.