Estimation of cyclic spectra using maximum likelihood filters

Conventional estimation methods for the cyclic spectra of cyclostationary processes are the temporally smoothed cyclic periodogram and the spectrally smoothed cyclic periodogram. In the case of short data records, both methods have low resolution and bad reliability. This paper uses maximum likelihood filters with modified analysis effective bandwidth to estimate cyclic spectra. Good performance in terms of resolution and reliability can be obtained using this method.