Source enumeration using a signal-selective information theoretic criterion

We propose a novel signal-selective approach to enumerating the signals of interest (SOIs) impinging upon an array of antennas. This new signal selective information theoretic criterion (SSITC) is derived by formulating the likelihood function for parameters of second-order cyclostationary signals in unknown stationary interference plus noise, and then applying Rissanen's MDL principle. We demonstrate through computer simulations that SSITC is able to perform correctly when there are more signals than sensors, the angular separations among the interferers and the signals of interest are arbitrarily small, and directional (spatially colored) noise is present. Thus, SSITC is an appropriate source enumerator to use in conjunction with cyclostationarity-exploiting direction-finding algorithms, such as cyclic MUSIC, and cyclostationarity-exploiting blind adaptive spatial filtering algorithms, such as SCORE and PCCA.