To　solve　the　limited　generalization　and　poor　adaptability　of　the　recognition　method　for　motor　imagery　electroencephalography　(MI-EEG),　the　traditional　growing　hierarchical　self-organizing　map　(GHSOM)　neural　network　is　improved,　and　an　adaptive　recognition　method　is　proposed　based　on　principal　component　analysis　(PCA)　and　improved　GHSOM　(IGHSOM)　neural　network.　The　hierarchy　growth　judgment　is　automatically　accomplished　according　to　the　quantization　error　of　the　expansion　neurons　in　upper　layer.　Thus,　IGHSOM　can　not　only　reflect　the　mapping　data　more　accurately　and　in　more　details,　but　also　improve　the　stability　and　adaptive　ability　of　the　network.　The　experiment　on　the　BCI　competition　data　set　was　conducted　to　assess　the　recognition　method;　the　PCA　was　used　to　extract　the　MI-EEG　features,　and　IGHSOM　was　employed　to　classify　the　features.　The　experiment　results　indicate　that　the　proposed　method　achieves　high　recognition　accuracy,　which　verifies　the　correction　and　effectiveness　of　the　improved　strategy　of　GHSOM　and　the　proposed　recognition　approach.　©,　2015,　Science　Press.　All　right　reserved.