Scholarlink

<p>To address the problem that traditional time-shifting phase-difference frequency estimation suffer from phase wrapping and reduced accuracy under large time-shift or noisy conditions, this paper proposes a frequency estimator combining Denoising Convolutional Residual Neural Network and phase difference. The proposed method formulates phase wrapping prediction as a classification task and employs a Denoising Convolutional Residual Neural Network to learn the nonlinear relationship between phase wrapping, noise, and time-shifting coefficients, enabling automatic unwrapping and precision compensation. Experimental results show that the proposed method maintains high accuracy and robustness under long time-shifting and low signal-to-noise ratio conditions, with the maximum absolute error below <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(5\times {10^{-4}}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>5</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>-</mo> <mn>4</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation> and the root mean square error decreasing monotonically with increasing time-shifting coefficients. The proposed approach effectively overcomes the limitations of traditional phase difference and provides a new solution for high-precision frequency estimation.</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

Frequency estimation using learned phase-wrapping compensation based on DNCNNRet and time-shift phase difference

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Ting Wang,
  • Wenjie Tan,
  • Zhenqing Luo,
  • Yuzhe Wang,
  • Yang Zhao

摘要

To address the problem that traditional time-shifting phase-difference frequency estimation suffer from phase wrapping and reduced accuracy under large time-shift or noisy conditions, this paper proposes a frequency estimator combining Denoising Convolutional Residual Neural Network and phase difference. The proposed method formulates phase wrapping prediction as a classification task and employs a Denoising Convolutional Residual Neural Network to learn the nonlinear relationship between phase wrapping, noise, and time-shifting coefficients, enabling automatic unwrapping and precision compensation. Experimental results show that the proposed method maintains high accuracy and robustness under long time-shifting and low signal-to-noise ratio conditions, with the maximum absolute error below \(5\times {10^{-4}}\) 5 × 10 - 4 and the root mean square error decreasing monotonically with increasing time-shifting coefficients. The proposed approach effectively overcomes the limitations of traditional phase difference and provides a new solution for high-precision frequency estimation.