1774 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 16, NO. 9, DECEMBER 1998 Adaptive Interference Cancellation for DS-CDMA Systems Using Neural Network Techniques Kaushik Das and Salvatore D. Morgera, Fellow, IEEE Abstract—The objective of this study is to apply and investigate a neural network-based decision feedback scheme for interference suppression in direct sequence code division multiple access (DS- CDMA) wireless networks. It is demonstrated that a decision feedback functional link equalizer (DFFLE) in combination with an eigenvector network can closely approximate a Bayesian receiver with significant advantages, such as improved bit-error ratio (BER) performance, adaptive operation, and single-user de- tection in multiuser environment. It is assumed that the spreading codes of the interfering users will be unknown to the receiver. This detector configuration is appropriate for downlink com- munication between a base station and a mobile user in a digital wireless network. The BER performance in the presence of interfering users is evaluated. The improved performance of such a DFFLE receiver for CDMA is attributed to the nonlinear decision boundary it evaluates for the desired user. The receiver structure is also capable of rapid adaptation in a dynamic communications scenario for which there is entry/exit of users and imperfect power control. The convergence performance and error propagation of the DFFLE receiver are also considered and exhibit reasonable promise for third generation wireless DS-CDMA networks. Index Terms— Adaptive interference suppression, neural net- works, single-user CDMA receiver, wireless networks. I. INTRODUCTION A critical issue for future wireless communications systems is the selection of appropriate multiple access techniques for reliable and affordable communication, anywhere and anytime. One type of wireless technology which has become popular over the last few years is direct-sequence code division multiple access (DS-CDMA). DS-CDMA, a form of a spread spectrum system, is capturing a significant interest in the personal communication services (PCS) research community. The attractive features of CDMA are its potential capacity increase over other multiple access methods, antimultipath capabilities, and its promise with respect to service integration in a bandwidth-on-demand fair sharing manner. There is considerable interest in developing advanced re- ceiver structures for future CDMA wireless systems. The perti- nent issues are complexity, near–far resistance, and robustness to the nonstationary multiple-access interference (MAI) en- vironment [3]. Since the conventional matched filter (MF) receiver suffers severe performance degradation in a near–far Manuscript received September 1997; revised February 1998. This work was supported in part under a Grant from the Canadian Institute for Telecom- munications Research. The authors are with the Information Networks and Systems Laboratory, Department of Electrical Engineering, McGill University, Montr´ eal, Quebec, H3A 2A7 Canada. Publisher Item Identifier S 0733-8716(98)08653-3. environment, there is an intensive research effort underway to develop receivers applying multiuser detection and/or in- terference cancellation techniques. The interest is also due to the fundamental fact that CDMA systems are limited by MAI and the belief that properly designed MAI-resistant receivers should certainly improve the performance and capacity. The optimal multiuser receiver structure was developed by Verd´ u [4], but its structure suffers from high implementation com- plexity. Recent efforts have focused on suboptimal receiver designs [5]–[8] in addition to an adaptive version, as pro- posed in [9]. Several articles [10]–[12] review the different approaches found in the literature. An important issue for these receivers is the degree of required knowledge of different parameters (time delay, signal strength of users, etc.) and/or signature sequences which are not a priori known and often difficult to estimate. Another important thrust in research in multiuser detection is the design of adaptive detectors, which self-tune the detector parameters from the observation of the received waveform [13]. The objective is to develop a single-user detector where the signature sequence of only one (desired) user is known. Here, detection is optimized in some sense [usually the min- imum mean-square (MMS) sense] for the multiuser channel. The receivers are known as MMSE receivers and are usually one shot, i.e., detection is based only on observation over one transmission interval. As a substitute for specific knowledge of the interfering users’ signature sequences, these detectors rely on adaptive signal processing and a training sequence. Linear MMSE receivers are proposed in [14]–[16] and a fractionally spaced DFE detector whose feedforward and feedback coeffi- cients are adapted to minimize mean-square error (MSE) can be found in [17]. Recently, an adaptive detector based on the principles of blind equalization has also been proposed and does not require training sequences for any user [18]. One promising approach to interference suppression is based on using neural networks. Aazhang et al. in [19] first reported on a study of a multilayer perceptron in this context and found that its performance is comparable to that of the optimum receiver. In [20], it is shown that the energy function of the recurrent neural network is identical to the likelihood function encountered in multiuser detection. The dynamics of the network are geared toward minimization of an energy function. The performance of the receiver is near optimum; however, the receiver is nonadaptive. Mitra et al. [21, [22] and Mulgrew [23] have investigated adaptive neural networks in the context of DS-CDMA demodulation. A single-layer perceptron is considered in [22] and is viewed as a least mean-square (LMS) 0733–8716/98$10.00  1998 IEEE