Several Typical Parameter Estimation Problems In Wireless Communication Systems

Over the last two decades, the field of wireless communications has been devel-oping at an explosive rate. The number of users has not only increased dramatically,but the business type and the amount of data transmission of each user are also growingfast. Therefore, increasing the capacity of the current communication systems is a keyproblem in wireless communications. It is well known that antenna arrays can providemore channel capacity for wireless communication systems than single antenna withoutany extra signal bandwidth or transmission power. They can also supply locating andtracking services to mobile terminals. These merits make the antenna array one of thekey technologies in the third and fourth generation of mobile communication systems.To reveal the potential of the antenna arrays, space-time signal processing is necessaryat the receiver which usually needs not only the information of the space-time wirelesschannel but also the accurate array manifold. This dissertation focuses on the uplinkspace-time wireless channel estimation problems and antenna array parameter calibra-tion problems in systems where the base stations are equipped with antenna arrays. Themain contributions of this thesis are shown as follows:1. In the uplink space-time wireless channel estimation problem, a space-time signalsubspace projection based channel estimation algorithm is proposed. Comparedwith the traditional signal subspace methods, the proposed method can better uti-lize the received signals corresponding to the user data. We formulate the channelestimation problem using a space-time signal model, then derive the space-timesignal subspace projection method in the context of maximum likelihood (ML)criterion. Furthermore, using the slowly-varying features in channels, we ex-tend the algorithm to obtain a multi-slot space-time signal subspace projectionalgorithm. In addition, the Cramer-Rao bound for channel estimation is derived.Simulation results show that the space-time signal subspace projection methodoutperforms the existing subspace projection methods.2. In the uplink space-time wireless channel estimation problem, a soft-based chan-nel estimation method is proposed by utilizing the soft bit information fed backfrom the decoder. The signal model is modified by formulating the differencesbetween the unknown user data and their soft information as additional Gaussiannoise. By using this modified model, an estimation problem based on the ML criterion is formulated. This ML estimation problem can be approximated eitherby a low complexity least square problem or by a semidefinite programming (S-DP) problem. In another method, we proposed to use the soft-based method andthe space-time signal subspace projection method in combination. At last, theCramer-Rao bound is derived for soft-based channel estimation methods. Sim-ulation results show that the proposed soft-based methods perform well and thecombination method works better such that the bit error rate (BER) of the turbosystem approaches the BER of the system using perfect channel state informa-tion.3. In the active array gain/phase calibration problem, an SDP method for uniformlinear arrays (ULAs) with known mutual coupling matrix (MCM) and a blockcoordinate descent (BCD) method for ULAs with unkown MCM are proposedbased on the ML criterion. We add the constraint that array gain/phase errorscan be upper bounded by a known value in practice. This bound constraint canimprove the calibration performance and make the calibration error smaller thanthe CRB without bound constraint at high noise levels.4. An ML direction of arrival (DOA) estimation algorithm is proposed for ULAswith unknown MCM. This algorithm obtains the DOA estimations by iterativelysolving the following two subproblems: DOA estimation problem with givenMCM and mutual coupling coefficients estimation problem with given DOA. Wesolve the first problem by a ML DOA estimation algorithm based on sum ofsquares (SOS) combined with SDP and the second problem by a semidefiniterelaxation (SDR) method. Based on the ML criterion, the proposed method canestimate the DOAs of coherent sources and approach the CRB.