One Pixel Image and RF Signal Based Split Learning for mmWave Received Power Prediction

TL;DR

This paper introduces a multimodal split learning framework that combines RF signals and minimal image data for accurate millimeter-wave received power prediction, enhancing privacy and communication efficiency.

Contribution

It presents a novel split learning architecture that effectively integrates RF signals and a single pixel image, improving prediction accuracy and privacy over RF-only methods.

Findings

Higher prediction accuracy with one pixel image and RF signals

Enhanced communication efficiency through payload compression

Maintained data privacy with minimal image data

Abstract

Focusing on the received power prediction of millimeter-wave (mmWave) radio-frequency (RF) signals, we propose a multimodal split learning (SL) framework that integrates RF received signal powers and depth-images observed by physically separated entities. To improve its communication efficiency while preserving data privacy, we propose an SL neural network architecture that compresses the communication payload, i.e., images. Compared to a baseline solely utilizing RF signals, numerical results show that SL integrating only one pixel image with RF signals achieves higher prediction accuracy while maximizing both communication efficiency and privacy guarantees.