Noise Power Density Estimation Based on Deep Learning Using Spectrograms Extracted from Wireless Signals

Myke D. M. Valadão, André L. A. da Costa, Éderson R. da Silva, Alexandre C. Mateus, Waldir S. S. Júnior

DOI: 10.14209/sbrt.2024.1571036528

Evento: XLII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2024)

Keywords: Noise Power Density Spectrogram Deep Learning

Abstract

In communication systems, noise is almost invariably present, originating from a multitude of sources and variables. These sources include thermal effects, interference, quantization, and channel imperfections, contributing to the random nature of noise. Determining noise levels is crucial and remains a pervasive challenge in communication systems, especially in recent times when better utilization of spectrum sensing is required. In this paper, we propose a noise prediction method based on deep learning using spectrograms extracted from wireless signals. The proposed method achieved promising results using several state-of-art computer vision architectures.

Download