The application of harmonic wavelet sub-band decomposition for reducing the Wigner-Ville Distribution cross-term in radar signal processing was demonstrated. The interaction between the signals as well as between the signal and noise were reduced by decomposing the multi-component signal into sub-bands, and then the separate components WVD were calculated, so as to obtain the original signal time-frequency distribution by using linear summation. The experiment results show that, for the multi-component signal with no cross-point in frequency domain, this method can effectively inhibit the cross-term and noise interference in WVD. The time-frequency resolving effect and the accuracy of instantaneous frequency extraction are superior to traditional WVD analysis methods. This work will help to improve the capacity of the radar signal detection and feature extraction.