Adaptive classification of underwater transients

This study is concerned with the classification of transient signals that can be represented by piecewise stationary processes. Within each stationary segment, the time series is modeled by a Gaussian autoregressive moving-average (ARMA) process. An algorithm for global classification, based on the entire transient, is presented. A global likelihood function, defined as the product of the generalized likelihood functions associated with each individual segment, performs the classification. For two arbitrary classes of transients, Monte Carlo simulations demonstrate that the method performs better than a classical classification scheme based on a single stationary segment