Scholarlink

<p>The increase in wireless radio applications required a large amount of spectrum to provide a high quality of services and continuous access. The scarcity of the spectrum is one of the serious concerns in radio applications. Spectrum sensing (SS) algorithms will play a critical role in cognitive radio network (CRN), and they will have the ability to access the spectrum to a greater extent. The utilisation of deep learning methods (DLM) such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) has huge potential to solve the spectrum issue. The effectiveness of various RNN architectures, including Long short-term memory (LSTM) and gated recurrent units (GRUs), in capturing long-term dependencies and enhancing spectrum sensing performance is discussed in the study. In this article, the RNNs and CNNs schemes are introduced to estimate and analyse the impact of the projected methods for the efficient detection of the spectrum without any interferences. The parameters such as probability of detection, probability of false alarm (pfa), bit error rate (BER), power spectrum density (PSD), and peak to average power ratio (PAPR) are evaluated and compared for the proposed and conventional schemes. The proposed algorithms detect the signal at low signal to noise ration (SNR) of 1.3&#xa0;dB and 1.7&#xa0;dB and have shown excellent pfa performance (4 and 7), SNR gain of 1&#xa0;dB to 8&#xa0;dB at the BER of 10<sup>–5</sup> is attained, and PAPR is optimised to 6.6&#xa0;dB with the optimal PSD values of -3010 and -2800. Hence, the experimental outcomes demonstrate that the projected CNNs and RNNs gave better performance than the conventional SS approaches.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spectrum sensing beyond 5G system: deep learning and conventional techniques analysis

  • 期刊论文

  • 发表日期：

    2025年1月24日

查看全文
  • Arun Kumar

摘要

The increase in wireless radio applications required a large amount of spectrum to provide a high quality of services and continuous access. The scarcity of the spectrum is one of the serious concerns in radio applications. Spectrum sensing (SS) algorithms will play a critical role in cognitive radio network (CRN), and they will have the ability to access the spectrum to a greater extent. The utilisation of deep learning methods (DLM) such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) has huge potential to solve the spectrum issue. The effectiveness of various RNN architectures, including Long short-term memory (LSTM) and gated recurrent units (GRUs), in capturing long-term dependencies and enhancing spectrum sensing performance is discussed in the study. In this article, the RNNs and CNNs schemes are introduced to estimate and analyse the impact of the projected methods for the efficient detection of the spectrum without any interferences. The parameters such as probability of detection, probability of false alarm (pfa), bit error rate (BER), power spectrum density (PSD), and peak to average power ratio (PAPR) are evaluated and compared for the proposed and conventional schemes. The proposed algorithms detect the signal at low signal to noise ration (SNR) of 1.3 dB and 1.7 dB and have shown excellent pfa performance (4 and 7), SNR gain of 1 dB to 8 dB at the BER of 10–5 is attained, and PAPR is optimised to 6.6 dB with the optimal PSD values of -3010 and -2800. Hence, the experimental outcomes demonstrate that the projected CNNs and RNNs gave better performance than the conventional SS approaches.