【8】Deep neural networks
MostcurrentspeechrecognitionsystemsusehiddenMarkovmodels(HMMs)todealwiththetemporalvariabilityofspeechandGaussianmixturemodels(GMMs)todeterminehowwelleachstateofeachHMMfitsaframeorashortwindowofframesofcoefficientsthatrepresentstheacousticinput.Analternativewaytoevaluatethefitis