Statistics for A new approach has been proposed for improving the performance of the Wigner-Ville distribution. This approach is based on signal decomposition and modified magnitude group delay function. Signal decomposition achieved by perfect reconstruction filter bank reduces significantly the existence of crossterms. The Gibbs ripple effect is due to truncation of the Wigner-Ville distribution kernel. The modified magnitude group delay function overcomes this effect without applying any window. Compared to those of Pseudo Wigner-Ville distribution and its versions, the proposed method has significantly improved performance in both time and frequency resolution as there is no time and frequency smoothing. Further, this method obeys better the desirable properties of time-frequency representation and has a better noise immunity. © 2003 Elsevier B.V. All rights reserved.