Channel Estimation Based on Machine Learning Paradigm for Spatial Modulation OFDM

TL;DR

This paper introduces a deep neural network-based detection scheme for SM-OFDM systems that reduces pilot overhead and cyclic prefix requirements, demonstrating superior performance over classical methods, with an ensemble network further enhancing robustness.

Contribution

The paper proposes a novel DNN-based end-to-end data detection method for SM-OFDM that implicitly handles channel estimation and introduces an ensemble network for better generalization.

Findings

DNN detection outperforms classical methods with less pilot overhead

Ensemble network improves generalization and performance

Significant reduction in cyclic prefix needed for reliable detection

Abstract

In this paper, deep neural network (DNN) is integrated with spatial modulation-orthogonal frequency division multiplexing (SM-OFDM) technique for end-to-end data detection over Rayleigh fading channel. This proposed system directly demodulates the received symbols, leaving the channel estimation done only implicitly. Furthermore, an ensemble network is also proposed for this system. Simulation results show that the proposed DNN detection scheme has a significant advantage over classical methods when the pilot overhead and cyclic prefix (CP) are reduced, owing to its ability to learn and adjust to complicated channel conditions. Finally, the ensemble network is shown to improve the generalization of the proposed scheme, while also showing a slight improvement in its performance.