Title: A LoRa signal denoising method based on deep learning
Authors: Baofeng Zhao; Shuo Feng
Addresses: College of Mining Engineering, Taiyuan University of Technology, Taiyuan, 030024, China ' College of Mining Engineering, Taiyuan University of Technology, Taiyuan, 030024, China
Abstract: In the actual industrial scene, given the low signal-to-noise ratio (SNR) and the coloured noise in the environment, the low-power wide-area network technology (LoRa) signal is easy to distort, and the transmission error rate is large. A LoRa denoising method based on deep learning is proposed. The noise is first extracted by feature learning by introducing a deep learning recurrent neural network. The learned model is used to suppress noise at the receiving end after the sampling of LoRa demodulation, and the transmission signal is optimised in the time domain and frequency domain to pre-mark possible error symbols. The simulation results show that compared with the traditional LoRa modulation and signal recognition method, the proposed method can greatly reduce the noise power of the received signal in a low SNR environment. The bit error rate is reduced by 20%-40%, and the transmission distance and receiver sensitivity are optimised.
Keywords: signal denoising; noise power; feature extraction; weak signal decoding.
DOI: 10.1504/IJSNET.2024.141611
International Journal of Sensor Networks, 2024 Vol.46 No.1, pp.15 - 23
Received: 22 Jan 2024
Accepted: 29 Jan 2024
Published online: 26 Sep 2024 *