Abstract: In the context of a small number of snapshots or unequal noise levels the noise eigenvalues of the covariance matrix are spreading, which results in the performance deterioration of the Akaike information criterion(AIC) and the Minimum Description Length(MDL). In this paper the Spatial Smoothing AIC(SSAIC) and Spatial Smoothing MDL(SSMDL) are presented. By spatial smoothing the spreading of the noise eigenvalues can be reduced remarkably, and hence the probability of correct detection can be increased, in addition, the consistency of the SSMDL is proved in detail. Finally, simulation results show that the SSAIC and SSMDL can improve the probability of correct detection remarkably.

Key words: array, parameter estimation, estimation of number of sources, spatial smoothing AIC, Spatial smoothing MDL