简介:Weconsiderinthispapertheproblemofrecursiveidentificationforstochasticsystemswhenthenoisemodeldoesnotsatisfythepositiverealconditionassociatedwithconvergenceofstandardalgorithms.Toavoidthepositiverealcondition,adaptivespectralfactorizationtechniquesareexploitedonthebasisofaclassofnon-standardtime-varyingrecursiveRiccatiequations.TheasymptoticpropertiesoftheRiccatiequationsarestudiedasacrucialsteptotheconvergenceresultsofthepaper.