Lamb　waves　are　widely　used　in　the　localization　and　quantification　of　structural　damage　for　plate-type　structure.　The　propagation　velocity　and　dispersion　characteristics　of　Lamb　waves　are　very　important　calculation　parameters.　The　method　to　estimate　the　dispersion　curves　of　Lamb　waves　is　proposed　based　on　K-means　clustering　algorithm.　Three　piezoelectric　transducers　with　different　center　frequencies　experimentally　transmit　the　Lamb　wave　signals,　and　laser　vibrometer　is　used　to　receive　signals　at　different　intervals　in　the　direction　of　wave　propagation.　The　phase　velocity　of　lamb　waves　is　estimated　by　many　pairs　of　adjacent　signals　peaks.　Use　Euclidean　distance　as　the　index　of　data　evaluation　to　find　the　data　center,　which　is　the　phase　velocity　to　be　solved.　And　the　dispersion　curves　are　further　estimated　and　inverted.　The　proposed　method　has　been　experimentally　verified　in　isotropic-aluminum　plate.　The　relative　error　of　the　actual　phase　velocity　of　lamb　waves　compared　with　all　measured　clustering　centers　is　less　than　1%,　and　the　overall　average　relative　error　is　less　than　0.4%.