In　order　to　make　full　use　of　the　redundant　and　noise　data　from　the　multi-resources,　enhance　the　identification　accuracy　of　structural　damage.　A　new　damage　identification　method　is　proposed　which　is　making　full　use　of　good　time-frequency　characteristic　of　the　wavelet　packet　and　data　fusion.　In　this　method,　the　response　signal　is　first　decomposed　so　as　to　extract　feature　parameters　by　wavelet　packet.　Then　different　feature　vectors　are　employed　to　identify　the　structural　damage　patterns　by　Euclidean　distance.　Finally,　weighted-average　is　employed　to　fuze　the　different　identification　results.　6　damage　patterns　of　a　7-story　steel　structure　are　presented　to　validate　this　method.　The　result　shows　that　this　proposed　method　can　enhance　the　identification　accuracy　greatly.