Parallel interference cancellation based on neural network in CDMA systems

In this letter, parallel interference cancellation (PIC) in code division multiple access (CDMA) was performed with two different structures by using a neural network (NN). In the first structure (receiver-1) the NN was used as a front-end stage of a one stage PIC circuit In the second structure (receiver-2), the NN was used instead of the one stage PIC circuit and it was trained as a multiple access interference (MAI) detector to perform the PIC process by subtracting the MAI from the outputs of the matched filter. The PIC is a classical technique in multi user detection process and its bit error rate (BER) performance is not good in one stage for most of the applications. For improving its BER performance, generally a multi stage PIC which has the high computational complexity is used. In this study, we have gotten a better BER performance than a three stages PIC receiver with both proposed receivers that have the lower computational complexity. Copyright © 2005 The Institute of Electronics, Information and Communication Engineers.