Study On Some Array Processing Methods For Spatial Signal Detection And Parameter Estimation

Signal detection and parameter estimation are two important subjects in advanced signal processing. They has attracted growing research interest in various fields, and have been applied widely, such as in wireless telecommunications, radar, sonar, image, audio, and seismology. In this dissertation, we investigate the array signal processing techniques to the multiuser detection and parameter estimation problem of far-filed or near-field sources of the multiple-input-multiple-output (MIMO) system, in which multiple transmit antennas and multiple receiver antennas are used. The primary contributions and original ideas included in this dissertation are summarized as below:1. The space-time coding technique is systematically addressed. Following the fundamentals of the space-time block code, the performance of differential space-time block code is studied under the spatial Rayleigh fading correlated channels, and a tight upper bound is also given on the symbol error rate.2.A novel blind multiuser detector is presented for frequency flat static channel, which uses adaptive Kalman filter to suppress multiple-access interference. Since this method employs differential space-time block decoding to restore the transmit symbol at the receiver, it does not require estimating the channel state information.3. Both the time-varying multipath fading of channels and multi-access interference degrade the performance of wireless communications, and render reliable date rate transmission a challenging issue. To this end, an adaptive multiuser receiver is proposed, which works efficiently as well in frequency selective fading Rayleigh channel. This receiver constructs a bank of decorrelating filters based on the minimum output energy (MOE) criteria. It extracts the signal energy of each path, and combines them by differential space-time decoder, achieving path diversity gain and spatial diversity gain.4. The existing high resolution approaches to DOA assume generally that theobservations are polluted by white Gaussian noise. When such an assumption does not hold in real applications, the performance deteriorates greatly. Based on the covari-ance matrix differencing approach, a fast estimation algorithm is presented, which can cope with the additive colored noise. Since this method does not proceed eigenvalue decomposition of transformed data matrix, its computational cost is relatively small. 5. Based on the fourth-order cumulants, a new range-frequency-DOA joint estimation algorithm is developed, which applies eigenvalues of a properly designed cumulant matrix to estimate the signal parameters. The three parameters of each near-field source are matched automatically, but the computational complexity is high due to eigenvalue decomposition of the high dimensional cumulant matrix. For this purpose, a modified algorithm is proposed, which applies the eigenvalues as well as the corresponding eigenvectors of two lower dimensional cumulant matrices. Both above two algorithms do not require searching for the spectral peak or pairing among parameters, and they have the power to resist an additive white or color Gaussian noise due to use of the fourth-order cumulants.