High-precision　spindle　bearing　is　one　of　the　most　critical　and　vulnerable　parts　in　a　motorized　spindle.　Its　unexpected　failure　may　lead　to　production　loss.　Stochastic　resonance　(SR)　is　a　weak　signal　detection　method,　which　can　obtain　noise　energy　in　strong　background　noise　and　enhance　incipient　fault　characteristics　of　spindle　bearing.　Based　on　the　fact　that　asymmetry　can　improve　the　enhancement　ability　of　asymmetric　bistable　SR　in　weak　feature　extraction,　we　introduce　an　underdamped　well-width　asymmetric　bistable　SR　(UABSR)　method　to　the　field　of　bearing　fault　diagnosis　for　the　first　time.　However,　the　engineering　application　of　UABSR　can　still　be　limited　by　two　aspects.　Firstly,　the　SNR　index　can　take　effect　only　when　the　actual　fault　frequency　is　obtained　in　advance,　so　the　UABSR　method　is　at　high-cost　in　real　practices.　Secondly,　an　appropriate　band-pass　filter　band　range　of　the　bearing　faults　can　hardly　be　obtained　due　to　the　massive　impulsive　noise　in　operations.　Here　an　improved　UABSR　method　for　spindle　bearing　fault　diagnosis　is　proposed.　Infogram　method　is　used　to　process　and　analysis　the　original　vibration　signal　for　resisting　the　influence　from　the　impulsive　noise　and　obtaining　more　accurate　frequency　range　of　spindle　bearing　fault.　In　addition,　time　domain　zero-crossing　(TDZC)　index,　as　the　index　of　the　improved　UABSR　method,　can　directly　reflect　the　fault　characteristics　of　spindle　bearings　without　knowing　the　accurate　fault　characteristic　frequency　in　advance.　Besides,　the　Quantum　Genetic　Algorithms　(QGAs)　and　the　fourth-order　Runge-Kutta　algorithm　are　combined　to　simultaneously　obtain　the　optimal　system　parameter,　the　asymmetric　ratio,　the　damping　factor　and　the　rescaling　factor　of　the　improved　UABSR　model.　Comparing　the　Infogram　and　original　UABSR　methods,　the　improved　UABSR　method　performs　better　effect　in　incipient　spindle　bearing　fault　diagnosis.　©　2013　IEEE.