arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1709.08243cs.SDeess.AS
收藏

一种混合DSP/深度学习的实时的全频段语音增强方法

A Hybrid DSP/Deep Learning Approach to Real-Time Full-Band Speech Enhancement

Mozilla Corporation(Mozilla公司);Mountain View, CA, USA(美国加利福尼亚州山景城(此处地址信息按要求不应提取,仅保留机构名Mozilla Corporation))

Jean-Marc Valin

摘要:

尽管噪声抑制是信号处理中的成熟领域,但它仍然高度依赖于估计器算法和参数的微调。在本文中,我们演示了一种混合DSP/深度学习的噪声抑制方法。具有四个隐藏层的深度神经网络用于估计理想的临界频带增益,而更传统的音调滤波器衰减音调谐波之间的噪声。该方法实现了显着更高的质量比传统的最小均方误差谱估计,同时保持足够低的复杂度,实时的【形】操作在48 kHz的低功耗处理器。

英文摘要:

Despite noise suppression being a mature area in signal processing, it remains highly dependent on fine tuning of estimator algorithms and parameters. In this paper, we demonstrate a hybrid DSP/deep learning approach to noise suppression. A deep neural network with four hidden layers is used to estimate ideal critical band gains, while a more traditional pitch filter attenuates noise between pitch harmonics. The approach achieves significantly higher quality than a traditional minimum mean squared error spectral estimator, while keeping the complexity low enough for real-time operation at 48 kHz on a low-power processor.

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。

↑