The problem of multisensor information fusion was studied in timefrequency domainand a new optimal criteria weighted by scalars was proposed.Wavelet transform was introduced in multiscale signal filtering.The state vector and observed value were decomposed into multiscale signalsand the multiscale signals (including approximate component and details) were updated.The local sensor estimation was fused via an optimal algorithm weighted by scalarsand timedomain state estimation was obtained through inverse transformation of wavelet.The method proposed can:1)complete multiscale filtering for the whole multisensor system;2)implement information fusion from the time domainand improve the estimation accuracy greatly;and 3)distribute fusion weights by scalarsthus only the computation of scalar weights is required and the calculation cost is reduced obviously.The simulation shows the effectiveness of the algorithm.