Application of Multiresolution Analysis to Direction-of-arrival Estimation

High-resolution methods of direction-of-arrival (DOA) estimation have been a topic of great importance in recent years for their wide-spread applications in radar, sonar, and mobile communication [7, 8]. Recently, most multiple source detection techniques are based on the eigenstructure decomposition of the covariance matrix. Among them, MUltiple SIgnal Classification (MUSIC) method [13] and Estimation of Signal Parameter via Rotational Invariance Technique (ESPRIT) [12] have been widely used for DOA estimation, harmonic analysis, frequency estimation, delay estimation, and the combinations thereof. Both approaches are implemented on the basis of a fixed, uniform, spatial sampling for all signal directions, which show the possibility of performance enhancement if a multiresolution signal representation is introduced [19]. Subband decomposition is a technique to split the spectrum of fullband signal with bandpass filters and subsampling to have a time-frequency decomposition, which provides a candidate to multiresolution method. Rao and Pealman proved in [11] that with subband decomposition, some superiorities can be obtained comparing with the direct estimation on the fullband signal. Classical wavelet transform (WT) [10] is a two-subband decomposition method, which presents the resolution both in frequency domain and time domain. Different mother wavelets are designed to deal with different scenarios. Wavelet packet (WP) is a much detailed WT by decomposing the approximations and details as well. Wavelet-based method for array processing has been studied by many researchers. The early successful application