Spectrum Sensing Using Wavelet Transforms and Filtering Under Signal Frequency Distortion and Fading Conditions

Abstract

This article explores improving the accuracy and reliability of spectrum sensing methods within cognitive radio networks. The primary focus is on how signal fading and frequency distortion influence the results of spectral analysis. These issues can severely impact the precision of signal detection, making adaptive methods and filters indispensable for accurately detecting changes in the spectral landscape. The purpose of the paper is to evaluate the effectiveness of various adaptive methods and filters – such as wavelet transforms, along with Butterworth, Chebyshev, and Kaiser filters—in improving the detection of changes within the spectral environment across different signal-to-noise ratio (SNR) levels. The research spans a broad frequency range, concentrating on pivotal technologies like 5G NR, Wi-Fi 6, DVB-T2, and GPS, each having unique requirements for signal precision and dependability. The spectrum sensing approach described in the article achieves high signal detection accuracy under favorable conditions, particularly when the SNR is strong. Experiments revealed that with SNR values above 1 dB, the signal detection accuracy (True Positive Rate, or TPR) for all technologies examined remains at or above 0.90. For instance, the TPR for 5G NR is 0.92 at an SNR of 1 dB, while for Wi-Fi 6, it stands at 0.90. However, the effectiveness of the method declines as the SNR decreases. For example, with 5G NR, the TPR drops to 0.70 at an SNR of -21 dB, indicating a heightened probability of false signal detection. Similar patterns are observed with Wi-Fi 6, where the TPR falls to 0.65, with DVB-T2 to 0.68, and GPS to 0.66. Additionally, the average noise level rises as SNR diminishes, making accurate signal detection increasingly challenging and emphasizing the need for further refinement of these methods. The findings underscore the need for ongoing advancements in spectrum monitoring, especially under low SNR conditions. Future research should prioritize developing new or refining existing adaptive algorithms capable of operating effectively in complex spectral environments. Exploring the impact of other filtering and transformation methods could also yield valuable insights. Moreover, the incorporation of machine learning techniques offers a promising path for boosting the adaptability and accuracy of spectrum monitoring in real-world telecommunication systems.