Here,　to　overcome　the　problem　of　using　traditional　Kalman　filtering　(KF)　algorithm　needing　known　measurement　noise　variance　and　process　noise　variance　and　assuming　they　being　constant　when　reconstructing　a　structure＇s　unmeasured　point　response,　a　structural　response　reconstruction　method　based　on　moving-window　Kalman　filtering　(MWKF)　algorithm　was　proposed.　The　feature　of　this　method　was　not　needing　to　set　variances　of　measurement　noise　and　process　noise　according　to　experience　in　advance.　Firstly,　the　moving-window　technique　was　used　to　estimate　variances　of　measurement　noise　and　process　noise　in　real　time.　Then,　based　on　KF　algorithm,　responses　of　structure　at　positions　without　sensors　were　reconstructed　by　using　response　information　of　limited　measuring　points.　Finally,　a　plane　single-span　frame　structure　was　taken　as　an　example　to　do　numerical　simulation　and　test　analysis.　The　analysis　results　showed　that　the　proposed　method　can　be　used　to　effectively　estimate　measurement　noise　variance　and　process　noise　variance　in　real　time;　the　reconstructed　dynamic　response　time　history　of　unmeasured　points　agrees　well　with　the　calculated　response　time　history　or　the　measured　one.　©　2021,　Editorial　Office　of　Journal　of　Vibration　and　Shock.　All　right　reserved.