Variational Probabilistic Method for Multiuser Detector over Wireless CDMA System

This work considers the new design of variational probabilistic method for code-divisionmultiple access CDMA system over unknown frequency-selective channels by decomposing of the observation data into a number of signal components. The author designs the virtual trellis model representing the ISI channel for each separating signal user. The design of virtual trellis is used to generate extrinsic probability in term of BCJR algorithm for exchanging with single channel decoder for turbo decoding as priori information without encoding. In addition, this paper proposed a variational probabilistic method framework to iteratively calculate a posterior probabilities (APPs) of each signal component for its symbol and channel parameters. In comparison with other similar receivers, the simulation results demonstrate superior performance over variety experiments. In current and next-generation wireless systems, direct sequence code-division multiple access (DS-CDMA) is the preferred ways to accommodate multiple users in a multipath fading environment. Various detection methods are reported in literature for uncoded CDMA system, such as maximumlikelihood (ML) detector, sphere decoding algorithm, zeroforcing detector (ZF), minimum-mean-square-error detector (MMSE) detector, and successive cancellation detector. For a coded system, on the other hand, iterative receiver designs can approach optimal performance with an affordable receiver complexity [I]. Capitalizing on powerful coding schemes, reliable transmission at very high data rates becomes possible. However, in order to exploit reliability, the receiver requires accurate estimates of both the channel and synchronization parameters to perform reliable data detection. An example of coding scheme application, iterative turbo multiuser detection [2] is based on a novel technique combining the linear minimum mean square error (LMMSE) criterion and soft interference cancellation (LMMSEIC) for computing extrinsic information feeding for individual channel decoders by iterative manner, assuming that the channel state information (CSI) is known to receiver. Iterative processing techniques based on turbo principles have been studied by many researchers [2][12]. When turbo principles are consider, there are two well known difference extrinsic information calculation are conceivable. First, linear minimum mean square error (LMMSE) criterion based extrinsic calculation [2]-[11]. Second, sequential monte carlo (SMC) based extrinsic calculation [12]-[13]. This paper proposes a new technique for extrinsic calculation and channel estimation based on minimum kullback-leibler (MKL) multiuser detector algorithm for joint iterative channel estimation and extrinsic information estimation by iterative fashion. To implement extrinsic probability calculation in term of BCJR algorithm [14] without encoding. Author model a new virtual trellis on a particular symbols modulation of kth user as discussion in next section. Computer simulations are employed to assess the performance of the proposed schemes compare with (LMMSEIC) iterative multiuser method. II. SYSTEM DESCRIPTION