MostcurrentspeechrecognitionsystemsusehiddenMarkovmodels(HMMs)todealwiththetemporalvariabilityofspeechandGaussianmixturemodels(GMMs)todeterminehowwelleachstateofeachHMMfitsaframeorashortwindowofframesofcoefficientsthatrepresentstheacousticinput.Analternativewaytoevaluatethefitis