每日 arXiv

音频 · 语音 · 音乐 · 声学

每日新爬取的 arXiv 论文，按评分排序。中文摘要帮你快速判断是否值得深读。

2026-08-21
日期2026-08-21
已评分—
均分—
最高—

Daily Papers — 2026-08-21

9 papers on audio, speech, music, and acoustics.

1. TurboBias 2.0: Streaming Context-Biasing for Production-Efficient ASR Systems

Authors: Vladimir Bataev, Lilit Grigoryan, Andrei Andrusenko, Nikolay Karpov, Vitaly Lavrukhin et al.

Categories: eess.AS, cs.AI, cs.CL, cs.LG, cs.SD Score: 7.30/10 (Obj:8 Id:6 Ind:8 Comp:8 Eff:8 Nov:5 Rep:8)

  • Strength: 在 C-Earnings22 上 Unified 流式模型（1.12s 延迟）经 per-stream global boosting + beam search 将 F-score 从 62.2 提到 87.5、WER 从 16.3% 降到 14.3%，且 decoder 级注册时 128 stream 批量解码速度与全局树相当（RTFx 1714 vs 1726），无需重训练或辅助 CTC 路径。
  • Weakness: 学术增量有限（三项均为自家解码栈的工程扩展），CTC-WS 对比建立在承认无法精确复现的基准发表值之上，且 3 小时内部医疗集不公开、无词表规模与温度敏感性扫描，限制边界验证。
  • Full review: Claude Code 全文七公理审稿

Contextualization is essential for production automatic speech recognition (ASR) systems, where user-provided phrases must be recognized accurately under strict latency constraints. Although many context-biasing methods improve recognition accuracy, they often do not address the practical requirements of modern production ASR systems: streaming inference, efficient batched decoding, user-specific context lists, and low runtime overhead. We propose TurboBias 2.0, a production-oriented framework for efficient phrase boosting in Transducer-based ASR systems. The framework extends GPU-accelerated TurboBias with a case-insensitive boosting graph and per-stream batched decoding, allowing each utterance in a batch to use an independent context-biasing configuration. This enables personalized context biasing for multiple simultaneous users without sharing or mixing their context lists. The proposed framework supports both offline and streaming inference and can be used with greedy and beam-search decoding. Experiments show that TurboBias 2.0 improves contextual phrase recognition while preserving low latency and high throughput.

2. Training DeepFilterNet with Accurate Room Acoustic Simulations Improves Single-Channel Speech Enhancement

Authors: Alessia Milo, Georg Götz, Steinar Guðjónsson, Daniel Gert Nielsen, Jesper Pedersen et al.

Categories: eess.AS, cs.LG, physics.comp-ph | 5 pages, 2 figures, IWAENC 2026 Score: 7.20/10 (Obj:1 Id:2 Ind:1 Comp:1 Eff:2 Nov:2 Rep:1)

  • Strength: 在 8 个训练配置上一致验证了高保真混合仿真 RIR 优于 DNS4 ISM RIR——pooled 配对差 PESQ +0.166、SI-SDRi +0.110（95% CI 严格为正），下游 ASR WER 相对改善 12.0%–23.8%（ISM 仅 0.5%–13.1%），全部在 unseen 实测 RIR 上评测。
  • Weakness: 核心 Hybrid RIR 数据集与训练代码、权重均未发布（仅开放评测代码），且整管线对比无单因素消融、无主观评测佐证，增量贡献无法与既有 MB-RIRs 等工作精确对账。
  • Full review: Claude Code 全文七公理审稿

We investigate how the realism of synthetic room impulse response (RIR) datasets affects the training of DeepFilterNet3 for single-channel speech enhancement. We compare a DNS4 image-source-method (ISM) RIR dataset with a higher-acoustic-fidelity dataset generated using hybrid wave-based and geometrical acoustics simulation. Rather than isolating individual simulation factors, we compare complete RIR generation pipelines while keeping the enhancement model unchanged. Models are evaluated on unseen measured RIRs using objective speech enhancement metrics and downstream automatic speech recognition (ASR). Training with the higher-fidelity dataset consistently yields modest improvements in objective metrics and substantially lower ASR word error rates than the ISM dataset. Although the experiments do not attribute these gains to individual modelling components, they show that increasing the overall realism of synthetic acoustic training data improves the generalization of DeepFilterNet3 to unseen measured environments.

3. DAMOS: Learning Distortion-Aware Speech Quality Assessment through Explicit Distortion Localization

Authors: Naiyuan Li, Li Dong, Diqun Yan

Categories: cs.SD, cs.AI Score: 6.30/10 (Obj:8 Id:7 Ind:8 Comp:5 Eff:6 Nov:6 Rep:4)

  • Strength: DAMOS 在 BVCC 上达话语级 SRCC 0.885、MSE 0.191，跨 7 个外部库平均 SRCC 0.746 超过 UTMOS（0.718），并构建了首个 15,000 条带帧级标注的部分失真合成数据集（定位 F1 0.807）。
  • Weakness: 头号卖点”显式失真定位”在推理期贡献近乎可忽略（掩码置零 SRCC 仅降 0.001，增益主要来自训练期），增益大头在 DSLA 层聚合而非定位机制，且数据集与代码权重均未发布、0 个开源仓库，可复现性缺失。
  • Full review: Claude Code 全文七公理审稿

Automatic speech quality assessment aims to predict Mean Opinion Scores (MOS) consistent with human subjective perception and is essential for evaluating speech generation, enhancement, and communication systems. For speech signals, especially synthetic speech, distortions often occur locally, and overall perceptual quality is usually dominated by a small number of perceptually salient distortion regions. However, most existing methods are primarily optimized with utterance-level MOS, which provides only coarse-grained supervision and offer no explicit indication of where perceptually important distortions occur. To address this limitation, we introduce explicit distortion localization as auxiliary knowledge for speech quality assessment. We construct the first partially distorted speech dataset with frame-level distortion annotations and train a localization model to generate distortion cues. Building on these cues, we propose DAMOS, a distortion-aware speech quality assessment framework that integrates localization information into the MOS prediction pipeline. Experiments on multiple public benchmarks demonstrate that DAMOS consistently outperforms existing methods and exhibits strong cross-dataset generalization, validating the effectiveness of explicit distortion localization for speech quality assessment.

4. μNet: Ultra-Low-Memory and Low-Complexity Speech Enhancement for Embedded Digital Signal Processors

Authors: Shrishti Saha Shetu, Jose Miguel Martinez Aponte, Nagashree K. S. Rao, Sharvin Vittappan, Oliver Thiergart et al.

Categories: eess.AS Score: 6.30/10 (Obj:8 Id:8 Ind:8 Comp:6 Eff:6 Nov:5 Rep:4)

  • Strength: μNet 以 90 KB 静态内存、28 MMACs、int8 全量化支持 4 ms 算法时延，主观 MUSHRA 77.78 超过 GTCRN 的 74.24，MSE 变体 BAK(MOS)=4.03 为对比方法最高。
  • Weakness: 方法层为 ULCNet 增量改造且代码、权重、训练脚本均未发布，SI-SDR 落后 GTCRN（13.61 vs 14.62 dB），4 ms 时主观分从 72.44 陡降至 57.87，低时延竞争力存疑。
  • Full review: Claude Code 全文七公理审稿

Speech enhancement on embedded digital signal processors (DSPs) imposes strict constraints on memory footprint, computational complexity, latency, and support for integer operations. Although recent DNN-based approaches have addressed these challenges individually, no unified framework in the literature simultaneously addresses all these requirements for practical deployment. In this work, we propose μNet, an ultra-low-memory, low-complexity, and low-latency end-to-end DNN model. The proposed method requires only $90$~KB of static memory and $28$~MMACs, while supporting an algorithmic latency as low as $4$~ms with performance comparable to state-of-the-art methods of similar complexity. Our experiments demonstrate that μNet is compatible with neural accelerators and supports full integer-arithmetic operations on consumer DSP platforms such as Cadence Tensilica HiFi 4/5.

5. Do SpeechLMs Hear Their Own Opinions? Diagnosing and Mitigating Previous-Belief Contamination in Streaming Emotion Understanding

Authors: Haoyue Liu, Zhichao Wang, Ye Chen, Haonan Deng, Xiaoying Tang

Categories: cs.SD, cs.AI Score: 6.20/10 (Obj:8 Id:6 Ind:6 Comp:8 Eff:8 Nov:5 Rep:2)

  • Strength: 反事实诊断量化了流式 SpeechLM 的自污染——固定音频仅替换注入的历史标签就使准确率从 72.50% 降至 30.42%、翻转 65.69% 的预测，EmoUpdate 在全部 8 个 model–benchmark 组合取得 S-BAcc 与 step accuracy 最优（最高 +69.71 与 +38.41），且训练免费、零额外调用。
  • Weakness: 论文声明发布全套复现 artifact 但正文无任何代码/数据链接，GitHub 检索零命中，两个自建基准无公开下载，第三方无法复现任何数字；效用对比全部框定在自建基线包络内，缺乏独立第三方 SOTA 的横评锚点。
  • Full review: Claude Code 全文七公理审稿

Streaming emotion understanding uses historical state while continuously interpreting current audio, often feeding the model’s previous prediction back as context. We show that this history conditioning can distort current perception. On a balanced CREMA-D-Stream counterfactual diagnostic, changing only the injected previous emotion label while holding the audio fixed reduces current-audio accuracy from 72.50% to 30.42% and flips 65.69% of predictions. The effect is strongly label-asymmetric, with prior pull ranging from 4.76% to 98.20%, revealing a failure we call previous-belief contamination (PBC). To address PBC, we introduce EmoUpdate, a training-free framework that separates current-audio perception from historical state revision through three components: (1) a prior-blind acoustic firewall that prevents historical state from entering perception; (2) an evidence-shrunk causal belief filter that introduces history only after observation formation and retains label-asymmetric transition structure only when supported by observed evidence; and (3) a closed-form decontamination operator derived from the same counterfactual measurements for serving stacks where firewalling is unavailable. Across four SpeechLMs and two streaming emotion benchmarks, EmoUpdate achieves the best step accuracy and state-balanced accuracy in all eight model–benchmark settings, improving S-BAcc by up to 69.71 points and step accuracy by up to 38.41 points over the strongest controlled baselines.

6. MusPyExpress: Extending MusPy with Enhanced Expression Text Support

Authors: Phillip Long, Hao-Wen Dong, Julian McAuley, Zachary Novack

Categories: cs.SD, cs.LG, eess.AS | Accepted at NeurIPS 2025 Workshop on AI for Music: Where Creativity Meets Computation; 10 pages, 6 figures Score: 6.20/10 (Obj:8 Id:6 Ind:8 Comp:6 Eff:6 Nov:5 Rep:7)

  • Strength: 用 PDMX 的 212,406 个含表情文本文件（95.33%）与 3.5M 条标记的量化统计证实了 MIDI 表示的缺口，并开源了 28 类注释对象的 MusPy 扩展（expressive 分支）+ 全套实验代码仓库。
  • Weakness: 生成实验全部复用 MMT tokenization 与既有 interleaving 方案，per-note perplexity 仅改善 2.80→2.64 且无任何配置全指标占优，全文没有人类聆听评测，也缺乏与 music21 的量化对比。
  • Full review: Claude Code 全文七公理审稿

Current work in modeling symbolic music primarily relies on representations extracted from MIDI-like data. While such formats allow for modeling symbolic music as sequences of notes, they omit the large space of symbolic annotations common in western sheet music broadly known as expression text, such as tempo or dynamics, which specify time- and velocity-dependent controls on the musical composition and performance. To alleviate this gap, we present MusPyExpress, an extension to the popular symbolic music processing library MusPy that enables the extraction of expression text along with symbolic music for downstream modeling. Utilizing this extension, we parse the PDMX dataset to illustrate the wealth of expression text available in MusicXML datasets. Additionally, we introduce multiple generative tasks, including joint expression-note generation, expression-conditioned music generation, and expression tagging, that take advantage of this additional notational information.

7. A Regularized Block Diagonal RLS Algorithm for Acoustic Echo Cancellation

Authors: Ruibin Hou, Chenggang Zhang, Yufeng Diao

Categories: cs.SD, eess.AS Score: 6.20/10 (Obj:1 Id:2 Ind:1 Comp:1 Eff:2 Nov:2 Rep:1)

  • Strength: RBD-RLS 将 RLS 的每步复杂度从 O(N²) 降至 O(NL)（N=512、L=32 时约 1/16 乘加量），在 100 次平均的白噪声/有色噪声/回声突变实验及 ICASSP AEC Challenge 真实场景中，其收敛与 ERLE 贴近 RLS/RLS-DCD，而基线 FRLS 在真实场景 0.5s 内完全发散。
  • Weakness: 方法为块对角近似与 Tikhonov 对角加载两个成熟技术的直接组合，未见系统对比分块/部分更新自适应滤波先例，无代码、无 MIS/ERLE 数值表与方差统计，真实场景仅单条语音 3 秒片段，且缺实测运行时间与相对 NLMS 工业路线的代价论证。
  • Full review: Claude Code 全文七公理审稿

While the recursive least square (RLS) algorithm is widely used in adaptive filtering applications like acoustic echo cancellation (AEC) due to its fast convergence rate, its high computational complexity severely limit its practical deployment for long filters. In this paper, a regularized block-diagonal RLS (RBD-RLS) algorithm is proposed to address these challenges. By approximating the inverse covariance matrix as a block-diagonal structure, RBD-RLS simplifies the update process into independent parallel computations of sub-blocks, effectively reducing the computational complexity. Additionally, Tikhonov regularization is applied to each sub-blocks for enhance numerical stability. A series of experimental results demonstrate that RBD-RLS maintains good convergence while significantly reducing computational complexity. Moreover, it still exhibits relative robustness in real-world scenarios.

8. SlimDiffuSE: Towards Efficient Diffusion-Based Speech Enhancement using Slimmable Networks

Authors: Nagashree K. S. Rao, Shrishti Saha Shetu, Mohamed Elminshawi, Emanuël A. P. Habets, Andreas Brendel

Categories: eess.AS Score: 5.53/10 (Obj:8 Id:5 Ind:5 Comp:5 Eff:6 Nov:4 Rep:5)

  • Strength: 在 DNS 2020 测试集上，用单参数集可瘦身网络将 SGMSE+ 的计算量从 125.40 降至 15.62 GMACs/Frame（降幅 87.5%），PESQ 仅从 2.85 微降至 2.77、SI-SDR 保持 16.7 dB。
  • Weakness: 仅单一非混响测试集、150 条样本且无主观听感评测，未与 CDiffuSE/FlowMSE/StoRM 等同类高效方案对比，87.5% 相对基线而非最优替代方案的增益待验证。
  • Full review: Claude Code 全文七公理审稿

Diffusion-based models are emerging in the speech enhancement domain and are achieving state-of-the-art performance across various benchmark datasets. A major downside of diffusion models is that data generation requires many evaluations of a typically large neural network, which results in high overall complexity. In this work, we propose a slimmable diffusion model that employs adaptive network widths throughout the data generation process to reduce computational cost. By using a greedy search algorithm to optimize the network width schedule, our method achieves performance comparable to baseline diffusion models with significantly reduced computational complexity. Notably, our approach reduces the computational complexity by up to $87.5\%$ without a significant drop in objective metrics, such as perceptual evaluation of speech quality (PESQ) and SI-SDR.

9. AudioWorldSim: Realistic Binaural Audio Datasets For World Models

Authors: Luis Vitor Zerkowski, Luiz Velho

Categories: cs.SD, cs.LG | 7 pages, 3 figures Score: 5.26/10 (Obj:7 Id:4 Ind:5 Comp:6 Eff:5 Nov:5 Rep:6)

  • Strength: 开源框架可批量产出连续双耳音频与逐动作标注（36 核 30 并行 worker、约 6 小时音频 5 小时内完成），并精确定位修复了 SoundSpaces 2.0 连续合成的点击伪影（IR 不等长致跨淡入淡出时间错位，零填充解决）。
  • Weakness: 核心「真实/无伪影」声明仅凭作者频谱图自证，无听感实验、无客观指标、无与任何基线（含原生 SoundSpaces 2.0）的定量对比，数据集因许可未发布，仓库无 license，全篇无下游模型验证。
  • Full review: Claude Code 全文七公理审稿

This technical report presents AudioWorldSim, an open-source platform designed to generate realistic binaural audio datasets and advance research in audio-based machine learning, particularly world models. Built as a custom extension of Meta’s SoundSpaces 2.0 platform, AudioWorldSim leverages their comprehensive acoustics framework, but focuses on the automatic rollout of random agent navigations, as well as implements crucial fixes to how continuous sound is composed. AudioWorldSim is made publicly available to the research community at https://github.com/Luizerko/AudioWorldSim to facilitate reproducibility.