Illegal and jamming signals disrupt wireless communication, causing degraded quality, false detection, and malfunctions. Spectrum monitoring is crucial across various fields like communications, broadcasting, radar, military, and security to detect these anomalies. Traditional detection algorithms struggle in diverse environments due to their parameter-dependent nature. Deep learning offers a promising solution with its ability to learn from large datasets and adapt to varying conditions. However, existing wireless datasets for deep learning are often proprietary or limited to narrowband single-signal modulation classification, unsuitable for wideband multi-signal spectrum monitoring. This paper introduces a novel spectrum monitoring dataset designed for object detection tasks. This dataset utilizes licensed frequency bands and employs Short Time Fourier Transform (STFT) spectrograms to capture both time and frequency characteristics. By annotating generated anomalous signals within these spectrograms, the dataset enables effective training of object detection models for identifying illegal and jamming signals in wideband spectrum monitoring applications.

This paper addresses the challenges of wideband signal beamforming in radar systems and proposes a new calibration method. Due to operating conditions, the frequency-dependent characteristics of the system can be changed, and the amplitude, phase, and time delay error can be generated. The proposed method is based on the concept of the sliding window algorithm for linear frequency modulated (LFM) signals. To calibrate the frequency-dependent errors from the transceiver and the time delay error from the true time delay elements, the proposed method utilizes the characteristic of the LFM signal. The LFM signal changes its frequency linearly with time, and the frequency domain characteristics of the hardware are presented in time. Therefore, by applying a matched filter to a part of the LFM signal, the frequency-dependent characteristics can be monitored and calibrated. The proposed method is compared with the conventional matched filter-based calibration results and verified by simulation results and beampatterns. Since the proposed method utilizes LFM signal as the calibration tone, the proposed method can be applied to any beamforming systems, not limited to LFM radars.

With the growing demand for wireless communication services, efficient anomaly detection in the electromagnetic spectrum is crucial to ensure service quality and prevent illegal spectrum usage. This study introduces a novel Synthetic Anomaly Detection (SAD)-based method tailored for spectrum anomaly detection. By leveraging synthetic anomaly data generation that captures the unique characteristics of spectrum signals, the proposed method addresses the limitations of conventional approaches, such as reconstruction errors and inadequate representation of spectrum anomalies. A rectangular mask and Gaussian noise were employed to simulate anomalies like tones, chirps, and pulses commonly found in LTE and 5G bands. The method achieved AUROC scores of 90% (image-level) and 82% (pixel-level) under ISR 0 dB, outperforming traditional unsupervised learning-based detection methods. An ablation study demonstrated the effectiveness of the synthetic anomaly generation process, highlighting its contribution to enhanced detection performance. These results validate the practicality and robustness of the proposed approach, providing a solid framework for spectrum anomaly detection in dynamic electromagnetic environments.

In modern society, LTE and 5G communications serve as essential infrastructures, making the stability of communication networks a critical challenge. However, advancements in wireless communication devices and technologies can lead to the transmission of illegal signals, device malfunctions, and radio interference, potentially causing degradation of communication performance. This study proposes a wideband spectrum monitoring method for LTE and 5G communication bands used in cellular networks by utilizing object detection techniques, thereby ensuring stable network performance and security. The proposed network accounts for the temporal and frequency characteristics present in signals on the spectrogram and is designed to exhibit high detection performance even for low-power signals. It achieved an mean Average Precision(mAP)50 performance of 80% at an Interference Signal Ratio(ISR) of –10dB corresponding to low power and an overall mAP50 performance of 93%.