Related Experiment Videos
Fast converging minimum probability of error neural network receivers for DS-CDMA communications
John D Matyjas1, Ioannis N Psaromiligkos, Stella N Batalama
1Department of Electrical Engineering, State University of New York at Buffalo, Buffalo, NY 14260, USA. matyjas@eng.buffalo.edu
IEEE Transactions on Neural Networks
|September 24, 2004
Summary
This study introduces a neural network receiver that improves direct-sequence code-division-multiple-access (DS-CDMA) performance by minimizing bit-error rate (BER). The novel adaptive algorithm enhances data recovery in noisy communication channels.
Area of Science:
- Electrical Engineering
- Computer Science
- Signal Processing
Background:
- Direct-sequence code-division-multiple-access (DS-CDMA) systems face challenges in accurately recovering user data due to interference and noise.
- Multilayer perceptron neural networks (NNs) offer potential for advanced signal processing in communication receivers.
- Optimizing receiver performance, particularly minimizing bit-error rate (BER), is crucial for reliable DS-CDMA communication.
Purpose of the Study:
- To develop and evaluate a novel neural network (NN) receiver architecture for enhanced information bit recovery in DS-CDMA systems.
- To introduce a fast-converging adaptive training algorithm designed to directly minimize the bit-error rate (BER) at the NN receiver output.
- To leverage specific properties of optimal decision boundaries and importance sampling (IS) within the adaptive learning process.
Main Methods:
- A multilayer perceptron neural network (NN) was designed as the receiver architecture.
- A novel adaptive training algorithm was developed, incorporating BER minimization directly into the learning process.
- The algorithm utilized constraints derived from optimum single-user decision boundaries for additive white Gaussian noise (AWGN) channels and embedded importance sampling (IS) principles.
Main Results:
- The proposed adaptive training algorithm demonstrated fast convergence.
- Simulation studies indicated significant improvements in bit-error rate (BER) performance for the NN receiver.
- The integration of BER, optimal decision boundary constraints, and importance sampling (IS) proved effective in receiver optimization.
Conclusions:
- The developed neural network receiver architecture with the adaptive training algorithm offers a promising approach for improving DS-CDMA system performance.
- The adaptive algorithm effectively minimizes BER by directly incorporating it into the learning process and utilizing channel-specific constraints.
- The study highlights the potential of advanced NN techniques and adaptive algorithms for robust data recovery in complex communication environments.