Fault　symptoms　of　running　gearboxes　must　be　detected　as　early　as　possible　to　avoid　serious　accidents.　Diverse　advanced　methods　are　developed　for　this　challenging　task.　However,　for　multiwavelet　transforms,　the　fixed　basis　functions　independent　of　the　input　dynamic　response　signals　will　possibly　reduce　the　accuracy　of　fault　diagnosis.　Meanwhile,　for　multiwavelet　denoising　technique,　the　universal　threshold　denoising　tends　to　overkill　important　but　weak　features　in　gear　fault　diagnosis.　To　overcome　the　shortcoming,　a　novel　method　incorporating　customized　(i.e.,　signal-based)　multiwavelet　lifting　schemes　with　sliding　window　denoising　is　proposed　in　this　paper.　On　the　basis　of　Hermite　spline　interpolation,　various　vector　prediction　and　update　operators　with　the　desirable　properties　of　biorthogonality,　symmetry,　short　support　and　vanishing　moments　are　constructed.　The　customized　lifting-based　multiwavelets　for　feature　matching　are　chosen　by　the　minimum　entropy　principle.　Due　to　the　periodic　characteristics　of　gearbox　vibration　signals,　sliding　window　denoising　favorable　to　retain　valuable　information　as　much　as　possible　is　employed　to　extract　and　identify　the　fault　features　in　gearbox　signals.　The　proposed　method　is　applied　to　simulation　experiments,　gear　fault　diagnosis　and　normal　gear　detection　to　testify　the　efficiency　and　reliability.　The　results　show　that　the　method　involving　the　selection　of　appropriate　basis　functions　and　the　proper　feature　extraction　technique　could　act　as　an　effective　and　promising　tool　for　gear　fault　detection.　(C)　2009　Elsevier　Ltd.　All　rights　reserved.