A novel wideband directional estimator of multiple periodic signals

Detection and classification of periodic signals is important in many application areas such as communication systems, medicine, radar, sonar, speech and the health monitoring of rotating machines. In general, it must be assumed that the profiles of replicated segments of periodic waveforms are unknown a-priori, that several periodic components may be present in the data and also that the sampling rate is not synchronised to the periodic replication rates. In this paper, we propose a novel maximum likelihood estimation algorithm that exploits two discriminants, periodicity and direction of arrival, to separate and detect multiple periodic wideband signals. Both periodicity and direction-of-arrival can be used simultaneously to reject noise and interference. The probability of detecting a signal at low SNR is then greatly increased and, additionally, periodicities which would otherwise be unresolvable become separable if they have different directions-of-arrival. Super-resolution is achievable by using a modified version of the incremental multi-parameter (IMP) algorithm. Many existing algorithms use Fourier or comb filter techniques to estimate one and sometimes more periodicities within a single time series. Our proposed algorithm operates directly in the time domain using a novel sample inter-leaving technique to avoid the sub-sample delays that are generally needed to coherently integrate asynchronously-sampled segments of a periodic waveform. We also show that our novel sample inter-leaving technique can also be exploited for wideband direction finding; where typically one would also require sub-sample delays