A novel approach for detecting the number of emitters in a cluster

Minimum description length (MDL) and Akaike information criteria (AIC) are two well-known techniques for estimating (or detecting) the number of sources observed by a direction finding array. The authors argue that nether MDL nor AIC optimizes detection performance. They do so by presenting an example algorithm applicable to a cluster of closely spaced sources that can usefully detect the number of emitters at a lower signal-to-noise ratio (SNR) than either MDL or AIC, or beamspace versions of MDL or AIC. The approach is not derived from information theoretic considerations as are MDL and AIC. A theoretical expression is presented for detection threshold SNR, which shows good agreement with simulation results.<>