Adaptive nonlinear least bit error-rate detection for symmetrical RBF beamforming.

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Rank-deficient multiple-antenna systems present challenges for nonlinear detection.
• Optimal Bayesian detection is computationally complex and sensitive to channel estimation errors.
• Existing linear detectors offer suboptimal performance in such systems.

Purpose of the Study:

• To propose a powerful symmetrical radial basis function (RBF) aided detector for rank-deficient multiple-antenna beamforming systems.
• To achieve near-optimal Bayesian detection performance using channel-impaired training data.
• To develop an adaptive training algorithm for the proposed RBF detector.

Main Methods:

• Exploiting the symmetry of the optimal Bayesian detection solution.
• Developing a novel nonlinear least bit error algorithm for adaptive training.
• Utilizing stochastic approximation to the Parzen window estimation for probability density function estimation.

Main Results:

• The proposed RBF detector approaches optimal Bayesian detection performance.
• The adaptive solution provides a signal-to-noise ratio gain exceeding 8 dB over linear benchmarks.
• Demonstrated performance improvements for configurations with 2 or 4 receive antennas.

Conclusions:

• The symmetrical RBF detector offers a robust and high-performance solution for nonlinear detection in challenging wireless systems.
• The adaptive training algorithm enables practical implementation with imperfect channel data.
• Significant performance gains are achievable, enhancing the efficiency of multiple-antenna systems.