A practical radial basis function equalizer.

Area of Science:

• Signal Processing
• Machine Learning
• Communications Engineering

Background:

• Radial Basis Function (RBF) equalizers offer advantages in non-linear channel equalization.
• Conventional RBF equalizer design requires a large number of basis function centers, increasing computational complexity.
• Efficient design of RBF equalizers is crucial for practical applications in digital communications.

Purpose of the Study:

• To develop a novel design process for radial basis function (RBF) equalizers that substantially reduces the number of basis function centers.
• To improve the efficiency and reduce the computational load of RBF equalizers without compromising performance.
• To demonstrate the effectiveness of the proposed RBF equalizer design through simulation studies.

Main Methods:

• A two-step algorithm for reducing the number of RBF centers: 1) selecting centers near the decision boundary, and 2) grouping and averaging selected centers.
• Estimation of channel order and delay using regression analysis to determine initial center requirements.
• Performance evaluation through simulation comparing the reduced-center RBF equalizer against a conventional RBF equalizer and a linear equalizer.

Main Results:

• Achieved a reduction in RBF centers by a factor of over 2000-to-1.
• The reduced-center RBF equalizer demonstrated performance comparable to the unreduced RBF equalizer.
• The proposed RBF equalizer significantly outperformed conventional linear equalizers in simulations.

Conclusions:

• The developed RBF equalizer design process effectively reduces the number of basis function centers while maintaining high performance.
• This method offers a computationally efficient alternative to traditional RBF equalizers.
• The optimized RBF equalizer presents a superior solution for channel equalization compared to conventional linear methods.