The　main　task　of　Cohen　class＇s　distribution　is　finding　suitable　kernel　function　to　reduce　the　cross-term　interference.　We　focus　on　adaptive　optimal-kernel　time-frequency　representation　which　adapts　its　kernel　function　with　time,　and　the　cross-term　interference　can　be　greatly　reduced　while　the　time-frequency　resolution　can　be　effectively　kept.　We　conduct　the　method　on　multi-component　synthetic　signal　and　compare　time-frequency　resolution　with　short-time　Fourier　transform　and　continuous　wavelet　transform.　Then　we　give　the　seismic　attenuation　characterizing　method　based　on　adaptive　optimal-kernel　time-frequency　representation.　The　3D　field　example　reveals　our　method　can　characterize　seismic　attenuation,　and　the　result　coincides　with　drilling　results　well.　©　2011　Society　of　Exploration　Geophysicists.