For several years now, the ITU-T's Perceptual Evaluation of Speech Quality (PESQ) has been the reference for objective speech quality assessment. It is widely deployed in commercial QoE measurement products, and it has been well studied in the literature. While PESQ does provide reasonably good correlation with subjective scores for VoIP applications, the algorithm itself is not usable in a real-time context, since it requires a reference signal, which is usually not available in normal conditions. In this paper we provide an alternative technique for estimating PESQ scores in a single-sided fashion, based on the Pseudo Subjective Quality Assessment (PSQA) technique.
Does a PESQNet (Loss) Require a Clean Reference Input? The Original PESQ Does, But ACR Listening Tests Don't
语音质量的感知评价(PESQ)需要一个干净的语音参考作为输入,但预测的结果(无参考)绝对类别评级(ACR)测试。在这项工作中,我们训练一个完全卷积的递归神经网络(FCRN)作为深度噪声抑制(DNS)模型,使用非侵入式或侵入式PESQNet,其中只有后者可以访问干净的语音参考。PESQNet被用作在DNS训练期间提供感知损失的中介,以最大化增强语音信号的PESQ分数。对于侵入式PESQNet,我们研究了两种拓扑结构,称为早期融合(EF)和中期融合(MF)PESQNet,并与非侵入式PESQNet进行比较,以评估和量化在DNS训练期间采用干净语音参考输入的好处。详细的分析表明,使用MF侵入式PESQNet训练的DNS分别比Interspeech 2021 DNS Challenge基线和使用MSE损失训练的相同DNS分别高出0.23和0.12个PESQ点。此外,我们可以证明,与使用非侵入式PESQNet训练的DNS相比,只获得了边际效益。因此,作为ACR听力测试,PESQNet不一定需要干净的语音参考输入,从而可以使用真实数据进行DNS训练。
MetricGAN+/-: Increasing Robustness of Noise Reduction on Unseen Data
The University of Sheffield(谢菲尔德大学)
语音增强系统的训练通常不包含人类感知的知识,因此可能导致不自然的声音结果。通过预测器网络将心理声学激励的语音感知度量作为模型训练的一部分最近引起了人们的兴趣。然而,这种预测器的性能受到出现在训练数据中的度量分数的分布的限制。在这项工作中,我们提出了MetricGAN+/-(MetricGAN+的扩展,一个这样的度量激励系统),它引入了一个额外的网络-一个“去生成器”,试图通过确保在训练中观察更广泛的度量分数来提高预测网络的鲁棒性(并通过扩展生成器)。在VoiceBank-DEMAND数据集上的实验结果显示,PESQ评分相对提高了3.8%(3.05 vs 3.22 PESQ评分),并且对看不见的噪声和语音有更好的泛化能力。
MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement
Research Center for Information Technology Innovation, Academia Sinica(中央研究院资讯科技创新研究中心);The Ohio State University(俄亥俄州立大学);Mila-Quebec AI Institute(米拉-魁北克人工智能研究所);National Institute of Information and Communications Technology(日本信息通信研究机构)
Single channel speech enhancement by colored spectrograms
University of Engineering and Technology, Peshawar(白沙瓦工程技术大学);College of Engineering, Qatar University(卡塔尔大学工程学院);National Center of Artificial Intelligence(国家人工智能中心)
本文提出基于彩色频谱图和 pix2pix GAN 的单通道语音增强方法,以极低计算成本取得与基线相当的 PESQ 和 STOI 提升。
Mamba-SEUNet: Mamba UNet for Monaural Speech Enhancement
College of Intelligence and Computing, Tianjin University(天津大学智能与计算学部);School of Electronic Information, Sichuan University(四川大学电子信息学院);Huiyan Technology (Tianjin) Co., Ltd(慧言科技(天津)有限公司);Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences(中国科学院深圳先进技术研究院)
Quality-Net: An End-to-End Non-intrusive Speech Quality Assessment Model based on BLSTM
Research Center for Information Technology Innovation, Academia Sinica(中央研究院资讯科技创新研究中心);National Taiwan University(国立台湾大学);Institute of Information Science, Academia Sinica(中央研究院资讯科学研究所)
DNN-based Source Enhancement to Increase Objective Sound Quality Assessment Score
NTT Media Intelligence Laboratories, NTT Corporation(NTT(日本电报电话公司)媒体情报实验室);University of Auckland(奥克兰大学);The University of Electro-Communications(电气通信大学)
LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement
Department of Electrical Engineering, National Taiwan University, Taiwan(台湾大学电子工程系);Research Center for Information Technology Innovation, Academia Sinica, Taiwan(中央研究院资讯科技创新研究中心);Center for Hearing Research, University of California Irvine, USA(加州大学尔湾分校听力研究中心)
CleanUMamba: A Compact Mamba Network for Speech Denoising using Channel Pruning
Faculty of Electrical Engineering, Mathematics and Computer Science, Delft University of Technology(电气工程、数学和计算机科学学院,代尔夫特理工大学);Leiden Institute of Advanced Computer Science (LIACS), Leiden University(先进计算机科学研究所(LIACS),莱顿大学)
PrimeK-Net: Multi-scale Spectral Learning via Group Prime-Kernel Convolutional Neural Networks for Single Channel Speech Enhancement
School of Electronic Information, Sichuan University(四川大学电子信息学院);College of Intelligence and Computing, Tianjin University(天津大学智能与计算学部);School of Medicine, Tohoku University(东北大学医学院);School of Computer Science and Engineering, Nanjing University of Science and Technology(南京理工大学计算机科学与工程学院);Department of Linguistics and Translation, City University of Hong Kong(香港城市大学语言学及翻译学系)