ZCLoRa: A Lightweight Deep Learning Based Receiver for Enhanced LoRa Decoding

doi:10.1109/LCOMM.2025.3544893

ZCLoRa: A Lightweight Deep Learning Based Receiver for Enhanced LoRa Decoding

2025-01-01

0

PRE

AI

DOI：10.1109/LCOMM.2025.3544893

原文链接

原文求助

摘要

En 中文

This letter introduces Zak-CNN LoRa (ZCLoRa), a novel receiver designed for low-power, long-range Internet of things (IoT) applications, addressing key limitations in conventional LoRa communication systems. In existing LoRa implementations, challenges such as low transmission power, unknown channel conditions, and degraded received signals lead to suboptimal decoding performance. To overcome these issues, we propose ZCLoRa, which leverages the Zak transform and exploits its sparse representation capabilities, enabling effective signal denoising. The Zak-transformed signal is processed by a convolutional neural network (CNN) for enhanced symbol decoding. We evaluate the performance of ZCLoRa in terms of symbol error rate (SER), showing significant improvements in decoding accuracy while maintaining the low complexity and power efficiency essential for IoT devices. Furthermore, we validate our model on USRPs, providing empirical evidence of its real-world effectiveness.

Keyword:

Transforms

LoRa

Symbols

Decoding

Fading channels

Noise

Internet of Things

Convolutional neural networks

Modulation

Signal to noise ratio

LoRa modulation

Zak transform

sparse representation

convolutional neural networks

USRP

ZCLoRa: A Lightweight Deep Learning Based Receiver for Enhanced LoRa Decoding

摘要

期刊

机构

引用论文