Spectrum Anomaly Detection: A Deep Learning Approach

Computer and Knowledge Engineering

مقالات آماده انتشار، پذیرفته شده، انتشار آنلاین از تاریخ 12 مهر 1404   

نوع مقاله: Special Issue

شناسه دیجیتال (DOI): 10.22067/cke.2025.91456.1144

نویسندگان

Motahareh Aghalari¹؛ Hossein Khaleghi Bizaki^{* 2}

¹Malek-Ashtar University of Technology, Tehran, Iran

²Malek Ashtar University of Technology

چکیده

The frequency spectrum is a limited and valuable resource and faces challenges due to anomalies. Hence, the detection of anomalies in the frequency spectrum is crucial for maintaining the integrity and reliability of telecommunication systems. These anomalies, which include jamming signals and interference, can disrupt communication channels and degrade system performance. This paper presents a comprehensive review of deep learning applications in spectrum anomaly detection, focusing on research conducted between 2017 and 2024. The review examines various pre-processing techniques used in spectrum anomaly detection, highlighting the widespread use of spectrogram and short-time Fourier transform (STFT), particularly in reconstruction-based methods, due to their effectiveness in capturing time-frequency information despite their computational challenges. Additionally, the study underscores the importance of selecting appropriate problem-solving approaches, such as classification, segmentation, or object detection, and tailoring models to suit specific tasks. These findings underscore the potential of deep learning-based approaches in enhancing spectrum monitoring and interference management.

کلیدواژه‌ها

Anomaly؛ Deep learning؛ Interference؛ Spectrum