Nonlinear demodulation and channel coding in EBPSK scheme.

Area of Science:

• Digital Communications
• Machine Learning in Signal Processing

Background:

• Extended Binary Phase Shift Keying (EBPSK) is efficient but faces challenges with conventional demodulation.
• Special Impacting Filters (SIF) improve Bit Error Rate (BER) but complicate posterior probability estimation for LDPC decoding.

Purpose of the Study:

• To reduce BER in EBPSK demodulation.
• To provide accurate posterior probability estimates (PPEs) for LDPC decoding.
• To introduce a novel nonlinear demodulation approach using Support Vector Machines (SVM).

Main Methods:

• Utilized Support Vector Machine (SVM) classifier for nonlinear demodulation.
• Selected a few sampling points from the SIF output for generating PPEs.
• Analyzed SVM performance against other methods and varying sampling rates.

Main Results:

• Achieved accurate posterior probability estimation using the SVM method.
• Significantly improved BER performance when applying LDPC codes with SVM-derived PPEs.
• Demonstrated SVM's advantage in adverse conditions and lower sensitivity to sampling rates.

Conclusions:

• SVM is an effective method for EBPSK demodulation and obtaining reliable PPEs for LDPC decoding.
• The proposed SVM approach offers superior performance and robustness compared to conventional methods.