Drunk　driving　is　a　serious　threat　to　traffic　safety.　But　there　are　many　serious　deficiencies　for　current　identification　methods,　which　are　based　on　drivers＇　physiological　parameters.　Operating　the　steering　wheel　belongs　to　continuous　driving　behavior　and　the　previous　research　reveals　that　the　steering　wheel　performance　is　closely　related　to　driver　state.　In　this　study,　steering　wheel　angle　signals　from　both　unimpaired　drivers　and　alcohol-impaired　drivers　were　collected　and　processed　through　wavelet　analysis.　Furthermore,　the　time-frequency　characteristics　of　the　signals　in　the　two　states　were　extracted.　The　results　show　that　the　frequency　of　adjusting　the　steering　wheel　was　significantly　higher　during　unimpaired　driving.　Conversely,　sudden　adjustment　of　steering　wheel　during　alcohol-impaired　driving　was　more　likely.　Based　on　the　coefficient　feature　of　wavelet　analysis,　the　standard　deviation　of　detailed　coefficients　of　wavelet　analysis　was　proposed　to　be　one　of　the　indices　for　identification.　The　validity　was　demonstrated　by　additional　experiments.　©　2011　ASCE.