Multi-signal time-frequency model fitting using an approximate maximum likelihood algorithm

By taking advantage of the moderate computational demand of the recently developed approximate maximum likelihood (AML) algorithms, it is shown that with algorithm refinement linear parametric model-fitting can be extended to a wide range of complex data interpretation tasks, including time-frequency and time-scale analyses. A novel version of the IMP (incremental multiparameter) AML algorithm (using a partially deterministic and partially stochastic signal model) is suggested for adaptive detection, tracking and extraction from time-series waveforms of several modulated signal components of differing bandwidths.<
>