Performance Analysis of ANN-Based Improved Modulation Classification for GFDM System

Abstract

With the introduction of new multicarrier modulation technologies, multicarrier waveform detection has grown more challenging and has become an open topic for current and future 5G/6G surveillance and signal interception. For multicarrier modulation, waveforms need to be recognized practically; for this, a practical recognition technique is required. Generalized Frequency Division Multiplexing (GFDM) results in the superimposition of multiple sub-symbols in the time domain, which leads to high non-linearity. One of the popular non-linearity reduction methods used in GFDM, ANN-NFFE (Artificial Neural Network-Nonlinear Feed Forward Equalizer), creates a variety of signal representation modulation matrices. The findings from the study show that GFDM behaves well in distinguishing several modulation approaches tested with various valid wireless channel deficiencies, including AWGN, multipath fading, and clock offset. The recommended ANN-NFFE design significantly enhances classification accuracy at low SNR values, which is suitable for practical cases. It achieves 86.1 % accuracy at −2 dB SNR, 96.5 % at 0 dB SNR, and 99.8 % at 10 dB SNR. The GFDM-based improved Proportional Fairness (PF) scheme achieved a throughput performance of 97.66 %, enhancing the overall efficacy of the GFDM system. Finally, the dispersion in the optical communication GFDM system is decreased using ANN-NFFE.