Applications to ocean acoustic data from a towed array and to speech p
rocessing are presented for an improved optimal time-domain beamformer
, which involves optimizing over all possible source bearings and time
series for multiple sources using simulated annealing. The convergenc
e of the parameter search is accelerated by accepting time series pert
urbations only when the energy decreases. A comparison with the conven
tional delay-and-sum beamformer illustrates that the optimal beamforme
r handles larger receiver spacing and larger source-to-receiver ratio.
Periodic ambiguities are essentially eliminated by using irregular re
ceiver spacing and the improved search algorithm. Weak sources are han
dled with fractional beamforming. Noise cancellation is possible if th
e parameters of the noise are included in the search space. Two-dimens
ional localization is performed for nearby sources.