The rapid damage estimation of building group in Gunkanjima is performed through the synchronized monitoring of MEMS acceleration sensor network. The anomalous behavior is detected by tracking the observed acceleration time series of each building. Then, the global damage of each building is investigated by singular value decomposition of the whole frame. Finally, the local damage within each building is scrutinized by evaluating the spectral ratios between observed points of the partial frames. This paper deals with the detection of anomalous behaviors related to structural damage from real sensor network data during two successive typhoons.
