Scholarlink

<p>In this paper, we provide a new version of two-sensor source separation recursive structure in sub-band form for acoustic noise reduction in hands-free telephony application, which is controlled by a new independent variable step-size normalized decorrelation method. This algorithm is used to reduce acoustic noise while improving voice quality, particularly in very acoustic noisy environments. This approach is known as the Independent Adaptive-step-size Subband Recursive Decorrelation (IA-SRD) algorithm. We conducted several simulations in a variety of acoustic noisy situations to evaluate the efficacy of the proposed version using numerous subjective and objective criteria, including time evolution of enhanced speech signals, mean square error (MSE), and segmental signal-to-noise ratio (Seg-SNR). Based on all obtained comparative results, the proposed algorithm demonstrates efficient performances when compared to basic decorrelation and sub-band decorrelation algorithms regarding of noise reduction level, convergence rate and voice quality enhancement.</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

New independent adaptive-step-size sub-band recursive decorrelation approach for noise reduction and speech enhancement

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Redha Bendoumia,
  • Ahcene Abed,
  • Islam Hassani,
  • Aissa Amrouche

摘要

In this paper, we provide a new version of two-sensor source separation recursive structure in sub-band form for acoustic noise reduction in hands-free telephony application, which is controlled by a new independent variable step-size normalized decorrelation method. This algorithm is used to reduce acoustic noise while improving voice quality, particularly in very acoustic noisy environments. This approach is known as the Independent Adaptive-step-size Subband Recursive Decorrelation (IA-SRD) algorithm. We conducted several simulations in a variety of acoustic noisy situations to evaluate the efficacy of the proposed version using numerous subjective and objective criteria, including time evolution of enhanced speech signals, mean square error (MSE), and segmental signal-to-noise ratio (Seg-SNR). Based on all obtained comparative results, the proposed algorithm demonstrates efficient performances when compared to basic decorrelation and sub-band decorrelation algorithms regarding of noise reduction level, convergence rate and voice quality enhancement.