On a generalization of maximum entropy method for processing of nonstationary multichannel complex valued data

Maximum entropy method for estimating power spectral density of time series has been generalized to incorporate the multichannel, complex, and nonstationary characteristics. A relation between the prediction-error-filter and the output has been derived and the prediction-error-filter are obtained by designing Wiener filter for multichannel nonstationary complex valued data. The computational procedure for a two point long prediction-error-filter has been described as an example.