arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.21141eess.AS
收藏

线性约束深度深度束成形器用于多说话人场景

Linearly Constrained Deep Beamformer for Multi-Speaker Scenarios

Faculty of Engineering, Bar-Ilan University(巴伊兰大学工程学院)

Ilai Zaidel, Ori Engel, Bar Engel, Sharon Gannot

AI总结:

本文提出了一种深度束成形框架,用于在多说话人环境中增强目标说话人。通过一种受增广拉格朗日框架启发的自适应多项损失,深度神经网络直接从噪声多通道输入中训练以估计波束成形权重,同时满足线性空间约束。该模型还受到目标相对传输函数和估计的干扰子空间的引导,能够将波束指向目标说话人并抑制干扰源,从而在整体增强性能上优于传统LCMV波束成形器。此外,与LCMV波束成形器相比,该模型产生的旁瓣更可控且背景噪声抑制更有效。

摘要:

我们提出了一种深度波束成形框架,用于在多说话人环境中增强目标说话人。深度神经网络(DNN)被训练以直接从噪声多通道输入中估计波束成形权重,同时通过受增广拉格朗日框架启发的自适应多项损失满足线性空间约束。该损失结合了信号重建与惩罚项,这些惩罚项强制对目标的无失真响应并抑制干扰子空间。模型进一步由目标相对传输函数(RTF)和估计的干扰子空间引导。所提出模型能够将波束指向目标说话人并抑制干扰源,从而在整体增强性能上优于传统LCMV波束成形器所构建的相同估计的空间签名。此外,与LCMV波束成形器相比,所提出模型产生的旁瓣更可控且背景噪声抑制更有效。

英文摘要:

We propose a deep beamforming framework for enhancing target speaker(s) in multi-speaker environments. A deep neural network (DNN) is trained to estimate beamforming weights directly from noisy multichannel inputs while satisfying linear spatial constraints through an adaptive multi-term loss with progressively increasing constraint weights. The loss combines signal reconstruction with penalties that enforce a distortionless response toward the target and suppress the interference subspace. The model is further guided by the target relative transfer function (RTF) and the estimated interference subspace. The proposed model can direct a beam toward the target speaker while directing nulls toward the interfering sources, achieving superior overall enhancement performance compared with the classical LCMV beamformer constructed by the same estimated spatial signatures. Furthermore, compared with the LCMV beamformer, the proposed model produces more controlled sidelobes and improved background-noise attenuation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑