Detection in Array Receiver Using Radial Basis Function Network

Detection of digital signals in the presence of interference and noise plays an important role in personal and mobile communication systems. The interference may arise from multipath propagation or from the multiple users accessing the system. In general, the detection problem can be formulated as a data classification problem. According to the classical detection theory, the optimal detector is provided by the Bayes hypothesis testing [l]. In practice, the statistical properties of the received data, such as the distribution function and the number of incoming signals, are unknown a priori. I t is of significant interest to investigate other non-statistical approaches. Traditionally, linear adaptive filters based on the least mean squares (LMS) and the recursive least squares (RLS) algorithms [2] are employed to combat the degradation due to the interference. They are suboptimal because they only generate hyperplanar decision boundaries in the observation space. Recently, the radial basis function (RBF) network has received a considerable amount of attention. It has the universal approximation ability [3] to construct robust non-linear decision boundaries. Besides, its massive parallelism and fast training time make it desirable for solving complicated tasks. In general, signals arrive at the receiver not only with different time delays, but also from different spatial angles. This spatial information cannot be exploited with a single antenna receiver, and is important in handling the scenarios where the h o m i n g signals are not time-delayed by multiples of a symbol duration. Recently, antenna arrays have attracted much attention in the framework of spatial diversity combining. In this paper, the RBF network is incorporated into an array receiver to solve the detection problem in the spatial domain. The RBF network is first reviewed. Employing the Bayes criterion as a benchmark, a decision-boundary comparison is then performed among the array receiving systems based on the RBF network, the LMS and the RLS adaptive filters. After that, simulation results are presented to compare the bit-error-rate (BER) performance of these array systems.