Auditorymodelhasbeenappliedtoseveralaspectsofspeechsignalprocessingfield,andappearstobeeffectiveinperformance.Thispaperpresentstheinversetransformofeachstageofonewidelyusedauditorymodel.Firstofallitisnecessarytoinvertcorrelogramandreconstructphaseinformationbyrepetitiousiterationsinordertogetauditory-nervefiringrate.ThenextstepistoobtainthenegativepartsofthesignalviathereverseprocessoftheHWR(HalfWaveRectification).Finallythefunctionsofinnerhaircell/synapsemodelandGammatonefiltershavetobeinverted.Thusthewholeauditorymodelinversionhasbeenachieved.Anapplicationofnoisyspeechenhancementbasedonauditorymodelinversionalgorithmisproposed.Manyexperimentsshowthatthismethodiseffectiveinreducingnoise.EspeciallywhenSNRofnoisyspeechislowitismoreeffectivethanothermethods.Thusthisauditorymodelinversionmethodgiveninthispaperisapplicabletospeechenhancementfield.
