Scholarlink

<p>This paper presents a new recursive double-sensor affine projection (RDSAP) algorithm powerful against acoustic noise and interferences for blind audio signals enhancement applications. The main idea behind the proposed algorithm is to introduce a fast converging algorithm with low computational complexity, denoising speech signals without introducing any perceptible distortion in its intelligibility, and improve the effectiveness of the Recursive Blind Source Separation structure. A detailed mathematic development of our proposed (RDSAP) algorithm carried out; as well, the ranges of the step-sizes are provided. The performance of the proposed RDSAP algorithm is compared with: (i) The Robust Dual Sensor Normalized Least Mean Square algorithm; The Robust Dual Sensor Fast Normalized Least Mean Square algorithm; and The Robust Dual Sensor Pseudo Affine projection algorithm. The comparison study is carried out in terms of the Convergence time, the Perceptual Evaluation of Speech Quality, the System Mismatch and finally, the output SNR values. The obtained results show the performances superiority of the proposed RDSAP algorithm. The research findings provide a new efficient tool for many fields of adaptive filtering, such as acoustic system identification, and blind audio signals separation.</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

Blind audio-denoising algorithm based on the double-sensor affine projection approach

  • 期刊论文

  • 发表日期：

    2025年3月10日

查看全文
  • Boumegouas Rahil,
  • Djendi Mohamed

摘要

This paper presents a new recursive double-sensor affine projection (RDSAP) algorithm powerful against acoustic noise and interferences for blind audio signals enhancement applications. The main idea behind the proposed algorithm is to introduce a fast converging algorithm with low computational complexity, denoising speech signals without introducing any perceptible distortion in its intelligibility, and improve the effectiveness of the Recursive Blind Source Separation structure. A detailed mathematic development of our proposed (RDSAP) algorithm carried out; as well, the ranges of the step-sizes are provided. The performance of the proposed RDSAP algorithm is compared with: (i) The Robust Dual Sensor Normalized Least Mean Square algorithm; The Robust Dual Sensor Fast Normalized Least Mean Square algorithm; and The Robust Dual Sensor Pseudo Affine projection algorithm. The comparison study is carried out in terms of the Convergence time, the Perceptual Evaluation of Speech Quality, the System Mismatch and finally, the output SNR values. The obtained results show the performances superiority of the proposed RDSAP algorithm. The research findings provide a new efficient tool for many fields of adaptive filtering, such as acoustic system identification, and blind audio signals separation.