Estimating number of sub-Gaussian emitters in a narrowband DOA estimation problem by using independent component analysis

Accurate determination of the number of emitters is an important and nontrivial problem in direction of arrival (DOA) estimation. The energy criterion based on singular values of the sampled data covariance matrix requires either a-priori knowledge of the signal-to-noise ratio (SNR) or the noise energy itself. More refined approaches, such as the Akaike information criterion (AIC) and the minimum description length (MDL) criterion fail when the signals are non-Gaussian. Thus, they are inapplicable to DOA estimation of communication signals, which generally tend to be non-Gaussian. The presented approach is based on independent component analysis (ICA). The information bearing source signals, obtained by blind source separation (BSS), are identified through measuring their distance from Gaussianity. A fixed threshold parameter in the kurtosis domain is used which can be set to accommodate a wide range of SNRs and data sample sizes.