Transform-Domain Statistical Signal Processing - a Wavelet-Based Approach

• Main Authors: CHU YI, 瞿怡
• Other Authors: Wen-Hsien Fang
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/60171650973730746901

博士 === 國立臺灣科技大學 === 電子工程技術研究所 === 86 === The recently introduced wavelets and wavelet transforms have received much interest in various facets of signal processing problems. In this thesis, we attempt to develop some wavelet-based statistical signal processing algorithms. The main contributions of this thesis are twofold. The first contribution of this thesis is the introduction of a Harr wavelet-based preprocessing scheme. This new scheme can be used in conjunction with the subspace-based one-dimensional and two-dimensional (2-D) frequency estimation algorithms to reduce the computational load involved. Moreover, we also introduce a new state space-based 2-D frequency estimation algorithm. The new algorithm utilizes two Hankel-block-Hankel-like auxiliary matrices, constituted by a novel partition-and-stacking of 2-D data, to fully exploit the subspace invariance property and handle all possible 2-D frequencies. The second contribution is the development of a novel narrowband partially adaptive beamformer with the generalized sidelobe canceller (GSC) as the underlying structure. The new beamformer employs the wavelet filters in the design of the blocking matrix of the GSC. This new M-band wavelet-based blocking matrix can block the desired signals from the lower path as required. In addition, the misadjustments and eigenvalue spreads of the covariance matrices of the blocking matrix outputs are decreased as compared with those of previous approaches. Moreover, a detailed performance analysis of this new GSC is carried out to facilitate the choices of the parameters involved, which include the regularity of wavelet filters and the number of M. To demonstrate the effectiveness of these new algorithms and related analysis, simulation results in various scenarios are also furnished. As compared with previous works, the new ones provide a more appealing tradeoff between performance and computational complexity.