The performance of bearing estimation using spatial domain forward–backward predictor with higher-order statistics
Chih‐Yuan Sung, Shiunn-Jang Chern · The Journal of the Acoustical Society of America · 1996
The modified forward–backward linear predictor (MFBLP) methods with higher-order statistics, viz., the second-order statistics and the fourth-order cumulants, together with the broadband array structure are developed in this paper for bearing estimation. Here, the desired source signals of interest are narrow band and the additive Gaussian noise sources in the related sensors are assumed to be spatially correlated or uncorrelated with each other and white/colored processes in the temporal domain. In this paper, the MFBLP method with second-order statistics for bearing estimation will be emphasized. Moreover, to extract the principal eigenvalues of presented MFBLP methods a new search algorithm is also proposed. An analytical study of the MFBLP method with second-order statistics is first developed, under the assumption that the additive noise is white Gaussian process. From the observation of analytic results, a modification of the MFBLP method with second-order statistics is suggested. From simulation results, it is shown that the MFBLP methods with higher-order statistics is superior to the conventional MFBLP method with the linear array data directly, in terms of threshold signal-to-noise ratio (SNR). This is especially true when data length becomes relatively larger. Moreover, the performance improvement of using the new search algorithm together with all the MFBLP methods is discussed thoroughly.