Blind equalization and multiuser detection of DS-CDMA signals in multipath channels.

摘要

Code division multiple access (CDMA) has quickly become the most promising technology in wireless communications.; In this dissertation, we use higher-order statistics (HOS) to develop blind equalization and detection methods for various CDMA systems. These include variable sequence length (VSL) CDMA, multicode (MC) CDMA, code-hopping (CH) CDMA, space-time block coded (STBC) CDMA, and CDMA packet radio networks.; We develop several modifications of the HOS-based inverse-filter method of Tugnait and Li [l2] to apply to VSL CDMA, MC CDMA, CH CDMA and STBC CDMA. The method of [12] is applicable to a single-rate short-code DS-CDMA system where the spreading code of each user is one symbol long and it repeats every symbol. In VSL CDMA system, users of different rates are accommodated by assigning spreading codes of different lengths. In MC CDMA, high-rate users are accommodated by subsampling and multiplying each subsampled stream onto different spreading codes. In CH CDMA, each user's spreading code is several symbols long which may be needed in military applications for covertness. In STBC CDMA (Alamouti's scheme [56]), a user's symbol sequence is coded and transmitted through two transmit antennas. We model all the above four CDMA systems as an equivalent basic single-rate multiple-input multiple-output (MIMO) model by modelling each user in the system as a combination of several “virtual users” with different spreading codes but shared channel and transmission delay. Several blind equalizers are designed to extract each virtual user. Our main contribution is the development and analysis of different forms of the penalty functions that are used in batch, iterative stochastic gradient algorithms designed to equalize and detect a desired user's information sequence.; A user identification and undesired user cancellation algorithm is also developed which can be used in CDMA packet radio network. CDMA packet radio is a combination of CDMA technology and local area network (LAN) where all users send data packets of information symbols which have been spread by a spreading code (as in CDMA) in a common time and frequency slot: Our algorithm is designed to identify and extract a desired user from packets from multiple transmitting users. Several user identification costs are developed to serve as test statistics in a binary hypothesis testing formulation (null hypothesis: the desired user is extracted; alternative hypothesis: the desired user is not extracted). (Abstract shortened by UMI.)