Abstract: A multiscale estimation method of signal based on multiple threshold processing is presented. First, the original measurement signal is decomposed with multiscale wavelet transform. Then the detail signals on each scale are further decomposed into sub-detail signals and sub-smoothing signals which are processedwith different thresholds respectively. The smoothing signals on the coarsest scale are processed with Kalman filtering and reconstructed with the detail signal on each scale. At last, the signal estimate value on the original scale is obtained.