Improved Approach for the Maximum Entropy Deconvolution Problem.

Area of Science:

• Digital Communications
• Signal Processing
• Information Theory

Background:

• The Gaussian probability density function (pdf) is commonly used for convolutional noise in blind adaptive deconvolution but is only accurate in later stages.
• Previous attempts using Edgeworth Expansion and Maximum Entropy for 16 Quadrature Amplitude Modulation (QAM) showed limited or parameter-dependent improvements.

Purpose of the Study:

• To develop a novel equalization method for blind adaptive deconvolution.
• To accurately model convolutional noise pdf for 16 QAM signals without requiring additional parameters.

Main Methods:

• Applied the Generalized Gaussian density (GGD) function and Edgeworth Expansion to approximate the convolutional noise pdf.
• Developed a new equalization algorithm based on this improved noise model.
• Evaluated performance using simulations for the hard channel case.

Main Results:

• The proposed method achieved improved equalization performance, particularly in terms of convergence time.
• Convergence was observed approximately 15,000 symbols faster compared to the original Maximum Entropy algorithm.
• The new approach eliminated the need for additional predefined parameters required by previous methods.

Conclusions:

• The Generalized Gaussian density function provides an effective model for convolutional noise in 16 QAM blind adaptive deconvolution.
• The novel equalization method based on GGD significantly reduces the number of symbols needed for reliable decision-making.
• This advancement offers a more efficient and practical solution for equalization in digital communication systems.