Abstract: The shortwave narrow-band interference in speech communication is time-varying so that it is more difficult to filter out for hard adaptive tracking. The article designed a kind of algorithm to resist time-varying interference based on discrete dyadic wavelet transform, which analysis the speech signal in time-frequency domain on the high resolution and keep the speech signal continuity. There are different characteristics in the relationship of wavelet transform coefficients and scale when the speech signal，the random noise and time-varying interference are transformed on different scales. In wavelet domain the speech signal is decompose, the time-varying interference is efficaciously concentrated to partial weighted vector. Consequently we could deal with signal and interference differently, to achieve good tracking of and filtering the time-varying interference. We take into account the posteriori SNR of the speech signal ,and also could deal with the serious interference、feeble interference、overlapped interference and non-overlapped interference differently, all of these are to depress voice distortion. Simulation results show that this algorithm is able to track the time-varying interference ， filter the time-varying interference and result out voice distortion smaller ,all of these are notch superior to the traditional algorithm.