Performance analysis of cyclostationary spectrum sensing with dynamic thresholding using artificial neural network under varying signal to noise ratio and noise variance conditions
Affiliations: [a]
Department of Electronics and Communication Engineering, Srinivasa Ramanujan Centre, Kumbakonam, Tamilnadu, India
| [b]
School of Electrical and Electronics Engineering, SASTRA Deemed to Be University, Thanjavur, Tamilnadu, India
Correspondence:
[*]
Corresponding author. M. Suba, Department of Electronics and Communication Engineering, Srinivasa Ramanujan Centre, Kumbakonam, 612001, Tamilnadu, India. E-mai: suba@src.sastra.edu.
Abstract: A key component of cognitive radio technology is spectrum sensing, which finds and accesses unused frequency bands to efficiently use the underutilized spectrum. A potential method for spectrum sensing called cyclostationary feature detection (CFD) uses the cyclostationary characteristics of signals to distinguish between the signal and noise. Artificial neural networks (ANNs) have been suggested in recent years as a method for CFD based spectrum detection, which increases detection accuracy and decreases complexity. However, the variable signal to noise ratio (SNR) and noise variance have an impact on the effectiveness of ANNs for CFD-based spectrum sensing. The effectiveness of ANNs for CFD based spectrum sensing under different SNR and noise variance conditions is evaluated in this work for the determination of threshold value in a dynamic way. We look into how SNR and noise variance affect the precision of probability of detection (Pd) and system complexity. Out analysis show how well ANNs work for CFD based spectrum detection with dynamic threshold value in the presence of changing SNR and noise variation. The findings demonstrate that ANNs may still obtain high Pd values with low SNR and large noise variance while maintaining a modest level of system complexity. According to our research, for a variety of SNR and noise variance situations, ANNs may be a viable option for CFD based spectrum detection in cognitive radio (CR) networks. The proposed approach can significantly improve the detection accuracy and reduce the complexity of the system, thereby enhancing the overall performance of cognitive radio networks. Based on the proposed work, it is determined that MPSK modulation function well with additive white Gaussian noise (AWGN), Rayleigh, and Rician channels up to a lower SNR value of – 30 dB and MQAM supports a lower SNR value of up to – 20 dB.
Keywords: Cyclostationary feature detection, ANN, varying SNR, noise variance, dynamic thresholding, probability of detection
