A new structural status feature extraction method based on data fusion and wavelet analysis is presented.The data measured by multi-sensors are fused by using an improved consensus algorithm.Itovercomes the shortcoming of the conventional consensus algorithm with two sensors,which has different confidence distances for different measurement precisions.The supporting matrix is fuzzified,which can avoid the subiective error in determining the threshold value.Structural eigenvectors extracted from the acceleration signals by data fusion and by no data fusion are compared through a numerical example.The result shows that the structural feature extracted by data fusion includes more structural status information in different sites and it can describe the structural status better.