Communication System with Walsh Transform-Based End-to-End Autoencoder

Abstract

This paper investigates the design of end-to-end (E2E) autoencoders within AI-enhanced communication systems. It emphasizes the advantages of transitioning from Fast Fourier Transform (FFT)-based Orthogonal Frequency Division Multiplexing (OFDM) to a modulation technique based on the Walsh–Hadamard transform (WHT). This study underscores the WHT’s use of aperiodic basis functions, in contrast with the periodic bases of Fourier transforms. The proposed E2E autoencoder model integrates neural networks in both the transmitter and receiver for signal processing. The model is trained to adapt the bit rate according to the measured channel signal-to-noise ratio (SNR) using the same neural network, enabling operation at low SNR levels (down to −10 dB). Additionally, the model was experimentally validated in a laboratory setting using a software-defined radio (SDR)-based system setup.