Brian M. Sadler
Army Res. Lab., Adelphi, MD 20783
Wideband array processing using a steered covariance matrix (STCM) [H. Wang and M. Kaveh, IEEE Trans. ASSP 33, 823--831 (1985)] was implemented and evaluated with experimental data from a small baseline aeroacoustic sensor array. The nonstationary character of acoustic signatures from maneuvering vehicles requires harmonic association and adaptive selection of an operating-frequency set for the STCM on a data-block by block basis. The sensor outputs are processed so that all selected frequencies with the same direction of arrival (DOA) have the same rank-one representation in the steered covariance matrix. This matrix was used in the steered minimum variance (STMV) method [J. Krolik, in Advances in Spectrum Analysis and Array Processing, Vol. II, Chap. 6 (Prentice-Hall, Englewood Cliffs, NJ, 1991)] to accomplish adaptive direction finding. Processing gain and corresponding improvement of accuracy in the DOA estimates was observed when compared to narrow-band subspace methods such as ESPRIT. The STMV DOA estimates are more stable and more reliable than the narrow-band methods. This is especially evident in the multivehicle case with a nearby vehicle much louder than more distant ones.