Study On Robust Array Parameters Estimation And Its Applications

Robust detection, localization and beamforming of spatial sources under non-diea conditions are studied in this thesis, and the applications of methods for the parameters estimation in array signal processing to the problems of carrier frequency estimation, synchronization and tracking for OFDM systems are also discussed In the chapter 2, we mainly discuss the spatial signal detection in color noise fields. Signal cyclostationarity can help us perform blind processing of DOA estimation and beamforming, the methods using signal cyclostationarity need the exact known cycle frequency, we deal with the robustness of cyclostationarity-based DOA estimation and beamforming. In the second part of this thesis, we extend the array parameters estimation approaches to the carrier frequency offset estimation, frequency synchronization and compensation in OFDM systems. Intensive researches focus on the estimation of integral part of carrier frequency offset, and a novel effective integral frequency offset estimation scheme is developed. Then, we discuss carrier offset estimation for intercarrier interference self-cancellation based OFDM, and the performance improvement of carrier frequency synchronization and tracking with receiver diversity. The details of this thesis are organized as follows.1. We prsents three novel approaches to estimate the number of sources impinging on an array of sensors. Firstly, we prove that the effect of small eigenvalues on the estimation of source numbers can be reduced by diagonally loading the corresponding correlation matrix, and the performance of the methods for determining the number of sources impinging on a sensor array can be improved. Secondly, by using linear interpolation, an appropriate cost function is constructed which avoid the use of the logarithm likelihood function that is lack of robust to the noise correlation. In applications with an arbitrary spatially correlated noise field, under the assumption that the sources are narrow-band and their temporal correlation is longer than that of noise, a new method to sources number estimation is proposed which combines the methods of instrumental variable and unitary transformation. Numerical computer simulations and real data tests show the validity of the proposed methods.2. Since the performance of cyclostationarity-based array processing methods suffers from severe degradation if there is a small mismatch in the cyclefrequency of the signals of interest. Two improved cyclic MUSIC algorithms for DOA estimation are introduced with which robustness to the cycle frequency error is emphasized. The problem of robust adaptive array beamforming for cyclostationary signals is also emphasized. We propose the efficient robust algorithm based on the conventional cyclic adaptive beamformig algorithm against the mismatch in the cycle frequency. The performance of the proposed techniques is compared with that of existing methods via simulation examples, which demonstrate a marked improvement in the robustness against imprecisely knowledge of the cycle frequency.3. Since the methodology in array parameters estimation can be applied to the carrier frequency estimation for OFDM systems, we disscuss the equivalence in two parameters estimation. An efficient method for estimating integral part of frequency offset is developed based on maximum likelihood principle, of which emphasis on the CFO ambiguity resolution. In this method, only one training symbol is required and the performance is superior to the tradional ones. Theoretic performance of estimation in integral part of CFO is derived and computer simulations are performed to show the superiority to the method available.4. We develop efficient maximum likelihood frequency synchronizers by utilizing the redundancy introduced by self-cancellation intercarrier interference scheme. The mean square error of the proposed synchronizer and the range of carrier frequency offset can be coped with are analyzed. Also, blind carrier frequency synchronization for OFDM systems with receive diversity is studied. Closed loops for OFDM frequency synchronication is developed with receive diversity, of which emphasis are the synchronization performance and identifiabiblity. Computer simulations evaluate the performance of the methods proposed with comparison to the Cramer-Rao bound.