The　structural　operating　conditions　may　significantly　differ　from　those　during　laboratory　tests,　where　the　structure　is　well　known,　well　installed　and　properly　excited.　For　structures　under　natural　loading　conditions,　or　excited　by　random　forces,　excitations　cannot　be　measured　and　are　usually　non-stationary.　Hence,　an　improved　operational　modal　analysis　is　a　useful　complement　to　the　traditional　modal　analysis　approach.　This　paper　presents　the　application　of　a　new　identification　procedure　for　structural　modal　parameters,　named　the　wavelet-based　identification　technique.　The　wavelet-based　identification　that　works　in　time-frequency　domain　is　used　to　identify　the　dynamic　characteristics　of　the　structural　system　in　terms　of　natural　frequencies,　damping　coefficients　and　modes.　It　is　shown　how　the　amplitude　and　the　phase　of　the　wavelet　transform　of　operational　vibration　measurements　are　related　to　the　eigenfrequencies　and　damping　coefficients,　and　the　wavelet-based　spectrum　analysis　is　employed　to　identify　the　mode　shapes　of　the　structure.　Two　examples　have　demonstrated　that　the　current　identification　results　are　in　accordance　with　those　previously　obtained　from　the　peak　pick　method　in　frequency　domain　and　the　stochastic　subspace　identification　in　time　domain.