Direction-of-arrival estimation for uncorrelated FM signals

A method of direction finding for FM signals based on averaged spatial time-frequency distributions is presented. This technique has been applied previously to algorithms such as MUSIC and maximum likelihood (ML). However, these estimation techniques do not take full advantage of the underlying diagonal structure of the source matrices. It is demonstrated that, by using the direction estimation for uncorrelated emitters (DEUCE) algorithm (Jansson, M. et al., IEEE Trans. Sig. Processing, vol.47, no.4, p.945-56, 1999), we can improve the performance in estimating the source directions. To use this approach, knowledge of the source auto-term locations on the time-frequency plane is required. In the absence of prior knowledge about the time-frequency signatures of the sources, an automatic point selection procedure is incorporated. The proposed scheme is demonstrated to yield results close to the optimal case of known auto-term locations.