TFF_aDCNN: A Pre-Trained Base Model for Intelligent Wideband Spectrum Sensing

In the beyond 5G (B5G)/6G era, to achieve ultra-dense and ultra-large-capacity intelligent connection of all things, an intelligent wideband spectrum sensing technology is particularly important. However, in an extremely wide frequency range, it is still a challenge to achieve high-precision and high-reconstruction-capability wideband spectrum sensing (WSS) under a very low SNR. We propose a Time-Frequency-Fused adjustable Deep Convolutional Neural Network (TFF_aDCNN). Meanwhile, a novel TFF_aDCNN-based sensing framework is also proposed. In this framework, we can obtain a pre-trained base model with a single distribution by training TFF_aDCNN. Then, for the sensing task in the actual environment, we use the base model for transfer learning, so that a newly trained sensing model can be obtained very quickly (i.e. fine-tuned model). In the TFF_aDCNN, we design a main network and an adjustable auxiliary network, where the former learns complex and abstract signal features, while the latter assists the main network in learning different data distribution patterns during the training process and regulates the focus direction of the main network during the perception process. Simulation results show that TFF_aDCNN can significantly reduce hardware cost and improve reconstruction accuracy and reconstruction capability, when compared with SOMP and SwSOMP-based WSS algorithms, single-dimensional deep learning spectrum sensing method, and deep learning-based WSS (DLWSS), especially at very low SNRs.

I. Introduction

Spectrum is an important and scarce strategic resource. With the rapid development of next-generation wireless communication technologies and the Internet of Things (IoT), the types and numbers of devices accessing wireless networks are exploding [1]. Therefore, to meet the spectrum requirements of various wideband devices, the future 6G communication technology will inevitably move into the ultra-wide spectrum range of millimeter wave and terahertz [2], [3]. However, the current static spectrum allocation strategy has made the contradiction increasingly acute between the low utilization of spectrum resources and the shortage of spectrum resources [4]. In the complex electromagnetic environment, the traditional Wideband Spectrum Sensing(WSS) method based on cognitive radio (CR) [5] is difficult to meet future performance requirements. Wideband spectrum sensing faces several challenges that narrowband spectrum sensing doesn't have. First of all, a wide range of spectrum requires super high-speed ADC to sample. Secondly, wideband spectrum sensing needs a large storage room for data. Although some methods can be used to achieve sub-Nyquist sampling, the signal will be contaminated with critical aliasing. Using sub-Nyquist sampling can reduce the requirement of a high-speed ADC, but the sampled signal is difficult to process. However, with the rapid development of artificial intelligence (AI) technology, it is a feasible solution to realize smart spectrum sensing in a very wide spectral range with the help of AI [6].