A hybrid discrete‐Wigner distribution/autoregressive moving average (DWD/ARMA) cyclespectrum estimator is developed and applied to signaturing of nonstationary acoustic returns. The ARMA effects a high resolution mapping from the time‐frequency Wigner plane to the cyclespectrum. The hybrid estimator is used to characterize cyclic correlations hidden in conventional spectral estimators that prove useful to signal source identification and to signal and noise decomposition. The ARMA performs the high resolution needed in the cycle‐frequency domain to characterize the signal source. The autoregressive model provides resolution, inclusion of the moving average into the model reduces noise sensitivity. The results of the hybrid DWD/ARMA cyclespectrum estimator are integrated into the development of a signature identification system and tested on sonar data. [Work supported by a NAVCOASTSYSCEN's Independent Research project sponsored by the ONR.]