Based　on　the　B-spline　wavelet　transform　and　BP　neural　network,　a　method　was　proposed　to　extract　and　recognize　the　features　of　transient　visual　evoked　potential　in　the　brain-computer　interface　system.　Based　on　the　analysis　of　brain　frequency　domain　mapping,　this　paper　carried　out　a　new　averaging　pre-treatment　method　to　Transient　visual　evoked　potential　(TVEP)　in　order　to　enhance　the　signal-noise　ratio;　Then,　based　on　the　B-spline　wavelet　transform　to　extract　the　features　and　design　a　BP　Neural　Network　Classifier;　At　last,　study　on　the　TVEP　data　collect　by　experiment,　obtain　a　higher　recognition　rate　and　verify　the　correctness　and　effectiveness　of　this　method.　©2010　IEEE.