Scholarlink

<p>Orthogonal Frequency Division Multiplexing (OFDM) is a cornerstone of modern wireless communication due to its spectral efficiency and inherent resilience to multipath fading. Despite these advantages, OFDM systems remain vulnerable to performance degradation caused by channel impairments, frequency offsets, and nonlinear noise, leading to a high Bit Error Rate (BER). This work proposes a deep learning-based enhancement to OFDM performance under multipath fading conditions. We implement a practical, end-to-end OFDM system on GNU Radio using two USRP X310 devices in an indoor environment. Instead of relying on built-in OFDM modules, we design and integrate system components individually, allowing for flexible signal processing. To mitigate channel effects and reduce BER, we explore two deep learning approaches: (i) Joint Channel Estimation and Symbol Detection (JCES), which simultaneously estimates the channel and detects symbols—even in the absence of a cyclic prefix—and (ii) a Cascaded Convolutional Neural Network (Cs-CNN) that explicitly estimates the channel response. Experimental results indicate that the suggested models are much better than the traditional ones like Least Squares (LS) and Minimum Mean Square Error (MMSE). The JCES model can reduce its BER by nearly an order of magnitude relative to LS estimation with a BER of around 10<sup>− 4</sup> at 22 dB SNR with 64 pilots and the Cs-CNN model can decrease BER performance by up to 30% over LS and approaches MMSE performance with Rician fading. These results illustrate the efficiency of deep learning to strengthen the robustness of OFDM, provide quantifiable benefits in terms of BER minimization and dependability, and indicate a new path in designing intelligent wireless communication systems.</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

Deep Learning Approaches for Channel Estimation and BER Reduction in Multipath Fading OFDM Systems

  • 期刊论文

  • 发表日期：

    2025年10月29日

查看全文
  • KiranKumar Humse,
  • C. Anjanappa

摘要

Orthogonal Frequency Division Multiplexing (OFDM) is a cornerstone of modern wireless communication due to its spectral efficiency and inherent resilience to multipath fading. Despite these advantages, OFDM systems remain vulnerable to performance degradation caused by channel impairments, frequency offsets, and nonlinear noise, leading to a high Bit Error Rate (BER). This work proposes a deep learning-based enhancement to OFDM performance under multipath fading conditions. We implement a practical, end-to-end OFDM system on GNU Radio using two USRP X310 devices in an indoor environment. Instead of relying on built-in OFDM modules, we design and integrate system components individually, allowing for flexible signal processing. To mitigate channel effects and reduce BER, we explore two deep learning approaches: (i) Joint Channel Estimation and Symbol Detection (JCES), which simultaneously estimates the channel and detects symbols—even in the absence of a cyclic prefix—and (ii) a Cascaded Convolutional Neural Network (Cs-CNN) that explicitly estimates the channel response. Experimental results indicate that the suggested models are much better than the traditional ones like Least Squares (LS) and Minimum Mean Square Error (MMSE). The JCES model can reduce its BER by nearly an order of magnitude relative to LS estimation with a BER of around 10− 4 at 22 dB SNR with 64 pilots and the Cs-CNN model can decrease BER performance by up to 30% over LS and approaches MMSE performance with Rician fading. These results illustrate the efficiency of deep learning to strengthen the robustness of OFDM, provide quantifiable benefits in terms of BER minimization and dependability, and indicate a new path in designing intelligent wireless communication systems.