NOISE SUBSPACE TECHNIQUES IN NON-GAUSSIAN NOISE USING CUMULANTS

We consider noise subspace methods for narrowband direetion-of-arrival or harmonic retrieval in colored linear non-Gaussian noise of unknown covariance and unknown distribution The non-Gaussian noise covariance is estimated via higher order cumulants and combined with correlation information to solve a generalized eigenvalue problem The estimated e igenvectors are used in a variety of noise subspace methods such as mu ltiple signal classification (MUSIC), MVDR, and eigenvector. The noise covariance estimates are obtained in the presence of the harmonic sig nals, obviating the need for noise-only training records. The covarian ce estimates may be obtained nonparametrically via cumulant projection s, or parametrically using autoregressive moving average (ARMA) models . An information theoretic criterion using higher order cumulants is p resented which may be used to simultaneously estimate the ARMA model o rder and parameters. Third and fourth-order cumulants are employed for asymmetric and symmetric probability density function (pdf) cases, re spectively. Simulation results show considerable improvement over conv entional methods,vith no prewhitening. The effects of prewhitening are particularly evident in the dominant eigenvalues, as revealed by sing ular value decomposition (SVD) analysis.