Sullivan E., Holmes J., Carey W.

Bearing estimation using an acoustically short towed array can be enhanced by incorporating the motion of the array into the signal model. It was previously theoretically shown (J. Acoust. Soc. Am., Vol. 115, No 4) that by casting the problem as a joint estimation problem, where the parameters to be estimated are the bearing and the fundamental frequency of the discrete Fourier transform of the broadband data, the bearing estimation performance of the synthetic aperture algorithm, as measured by the variance of the bearing estimate, can exceed that of the conventional array processor. This occurs because the processor is able to exploit the bearing information contained in the Doppler. A Kalman filter is utilized to carry out the joint estimation problem. The two unknowns, the fundamental source frequency and the bearing angle, constitute the state vector elements. The measurement model is the frequency averaged phases of the Fourier components. The data are preprocessed by computing a sequence of overlapped Fourier transforms, and averaging the phases of the frequency lines. This sequence of averaged phases then constitutes the time series which is the input to the Kalman filter. An auxiliary measurement equation is used to relate the observed fundamental frequency to that of the source frequency. After a short theoretical introduction, experimental results based on the radiated noise of a ferry under way in Nantucket sound are show. Results of successful bearing estimation with an array of acoustic length of less than one wavelength will be compared to results using the same data but with a conventional beamformer. The improvement in performance of the synthetic aperture algorithm over the conventional algorithm will be quantified.