Abstract:Recently, compressive sensing (CS) and spectrum sensing have been two hot topics in the signal processing and cognitive radio network (CRN) fields, respectively. Due to the sampling rate limitation of the analog-to-digital converter in spectrum-sensing circuits, some works have proposed integrating these two techniques to achieve low-overhead spectrum sensing in CRNs. These works aim to minimize spectrum reconstruction errors based on linear regression methods, and ℓ1-norm is typically used to make a tradeoff between spectrum sparseness and reconstruction accuracy. However, since the interference range of primary users is limited, multiple clusters in the CRN may not share a common sparse spectrum, and thus, the ℓ1-norm may not be appropriate to handle all clusters in CS inversion. Hence, we propose a novel multitask spectrum-sensing method based on spatiotemporal data mining methods. In each cluster, we assume that the spectrum sensing is executed in a synchronized way. The cluster head (CH) manages the operations, and a common sparseness hyperparameter is used to make a consensus decision. Among multiple clusters, synchronized CS sampling is not required in our scheme; instead, the Dirichlet process prior is employed to make an automatic grouping of the spectrum-sensing results among different clusters with a common sparseness hyperparameter shared inside each group. To exploit the time-domain relevance among consecutive CS observations, a hidden Markov model is employed to describe the relationship between the hidden subcarrier states and the consecutive CS observations, and the Viterbi algorithm is used to make an accurate spectrum decision for each secondary user. Simulation results show that our proposed algorithm can successfully exploit the spatiotemporal relationship to achieve higher spectrum-sensing performance in terms of normalized mean square error, probability of correct detection, and probability of false alarm,- compared with a few other related works.

Elimination of Index Ambiguity for Overlapped Signals in Spectrum Sensing

QTFN: A General End-to-End Time-Frequency Network to Reveal the Time-Varying Signatures of the Time Series

TF(2)AN: A Temporal-Frequency Fusion Attention Network for Spectrum Energy Level Prediction

Radar Target Detection Based on OTFS Signaling: A Deep Learning Approach

Boost Spectrum Prediction with Temporal-Frequency Fusion Network Via Transfer Learning

Local maximum synchrosqueezing adaptive transformation for cross-instantaneous frequencies analysis

Unified Sparse Time–Frequency Analysis: Decomposition, Transformation, and Reassignment

Federated Spectrum Learning for Reconfigurable Intelligent Surfaces-Aided Wireless Edge Networks

A Novel Underdetermined Blind Source Separation Algorithm of Frequency-hopping Signals via Time-frequency Analysis

Multiple Sound Source Counting and Localization Based on TF-Wise Spatial Spectrum Clustering

A Detecting Algorithm Based on CAF for Overlapping Frequency-Spectrum Multi-Signals

Enhanced Micro-Doppler Feature Extraction Using Adaptive Short-Time Kernel-Based Sparse Time-Frequency Distribution

Sparse Time-Frequency Decomposition for Multiple Signals with Same Frequencies.

Sub-Nyquist Spectrum Sensing and DOA Estimation With Space-Time Trilinear Modeling

Multitask Spectrum Sensing in Cognitive Radio Networks via Spatiotemporal Data Mining

Deep Learning Aided Time-Frequency Analysis Filter Framework for Suppressing Ionosphere Clutter

Time-Frequency Energy Sensing of Communication Signals and Its Application in Co-Channel Interference Suppression

Kernel Design of bilinear TFD for multiple LFM signals with unbalanced amplitudes

Transient Interference Suppression Method Based on an Improved TK Energy Operator and Fuzzy Reasoning

Multiple Signal TDOA/FDOA Joint Estimation with Coherent Integration

An efficient cooperative spectrum sensing algorithm in time domain