BLIND DECONVOLUTION OF COLORED SIGNALS BASED ON HIGHER-ORDER CEPSTRA AND DATA FUSION

A new nonparametric blind deconvolution algorithm for coloured non-Gau ssian signals is presented. The goal of blind deconvolution is the rec onstruction of the input of an unknown linear time-invariant (LTI) mix ed-phase system based only on the system output. The existing deconvol ution schemes that require the least amount of knowledge about the inp ut signal and the LTI system were developed for white input signals, o r rely on parametric modelling of both the system and the input. To de velop nonparametric algorithms for the deconvolution of coloured proce sses of unknown statistics, we are forced to consider a two-channel ap proach. The proposed algorithm utilises the data collected by two diff erent receivers, each being the output of a different system due to th e same input. The two systems are then reconstructed combining higher- order statistics of the measured signals and the theory of signal reco nstruction from the higher-order spectral phase only.