Considering the sparseness of Gabor coefficients of the underlying signals in joint time-frequency plane
this paper describes an adaptive signal recovery approach
obtained by shrinking the Gabor coefficients based on Bayes rule.Via the assumption of prior distributions of Gabor coefficients for the unknown signals and noises
we present a closed form expression for the posterior estimation of Gabor coefficients of the unknown signals
implementing time-variable filtering in time-frequency domain.The validity and effectiveness of the proposed scheme are demonstrated with numerical experiments.