论文解读
SPAM: Style Prompt Adherence Metric for Prompt-Based TTS
语音合成 | 7.0/10
对比学习
论文解读
Spatial-CLAP: Learning Spatially-Aware Audio–Text Embeddings for Multi-Source Conditions
空间音频 | 8.5/10
论文解读
Speech Emotion Recognition based on Hierarchical Transformer with Shifted Windows
语音情感识别 | 8.0/10
论文解读
SpeechCT-CLIP: Distilling Text-Image Knowledge to Speech for Voice-Native Multimodal CT Analysis
医疗AI | 7.5/10
论文解读
Style-Disentangled Diffusion for Controllable and Identity-Generalized Speech-Driven Body Motion Generation
语音驱动动作生成 | 7.0/10
论文解读
SynaSpot: A Lightweight, Streaming Multi-modal Framework for Keyword Spotting with Audio-Text Synergy
关键词检测 | 7.5/10
论文解读
Temporally Heterogeneous Graph Contrastive Learning for Multimodal Acoustic Event Classification
音频事件检测 | 8.5/10
论文解读
The Curious Case of Visual Grounding: Different Effects for Speech-and Text-Based Language Encoders
模型评估 | 8.0/10
论文解读
Towards Effective Negation Modeling in Joint Audio-Text Models for Music
音乐理解 | 7.5/10
论文解读
TTA: Transcribe, Translate and Alignment for Cross-Lingual Speech Representation
语音识别 | 7.5/10
论文解读
WavLink: Compact Audio–Text Embeddings with a Global Whisper Token
音频检索 | 8.0/10
论文解读
Robust Audio-Text Retrieval via Cross-Modal Attention and Hybrid Loss
音频检索 | 7.5/10
论文解读
ATRIE: Adaptive Tuning for Robust Inference and Emotion in Persona-Driven Speech Synthesis
语音合成 | 7.0/10
论文解读
Materialistic RIR: Material Conditioned Realistic RIR Generation
音频生成 | 7.5/10
论文解读
ATIR: Towards Audio-Text Interleaved Contextual Retrieval
这篇论文旨在解决现有音频-文本检索方法无法处理查询和文档中音频与文本交错出现（如多轮对话、混合输入）的局限性。为此，作者定义了音频-文本交错上下文检索（ATIR）任务，并构建了一个包含约8.8万对样本的大规模基准。为解决直接应用多模态大语言模型（MLLM）时音频token冗余导致的效率和精度问题，论
论文解读
Deep Supervised Contrastive Learning of Pitch Contours for Robust Pitch Accent Classification in Seoul Korean
这篇论文旨在解决将连续变化的基频（F0）曲线映射到首尔韩语中离散、不变的音高重音类别（如LHLH, HHLH）这一难题。传统方法易受F0测量噪声和说话人差异的影响。为此，作者提出了**Dual-Glob**，一个深度监督对比学习框架。其核心是通过一个**双分支（干净视图和增强视图）编码器**，在共享
每日研究速递
语音/音乐/音频论文速递 2026-04-22
共分析 21 篇语音/AI 论文
论文解读
Joint-Centric Dual Contrastive Alignment with Structure-Preserving and Information-Balanced Regularization
这篇论文旨在解决音频-文本多模态表示学习中的一个关键挑战：如何在低资源、长序列且模态维度严重不平衡（音频高维、文本低维）的情况下，实现有效的跨模态对齐，同时保留各自的特异性信息。为此，作者提出了HILBERT框架。该方法首先利用冻结的预训练音频（如HuBERT）和文本（如T5）编码器提取片段级特征，
论文解读
ProSDD: Learning Prosodic Representations for Speech Deepfake Detection against Expressive and Emotional Attacks
这篇论文旨在解决当前语音深度伪造检测（SDD）系统在面对富有表现力和情感的合成语音攻击时泛化能力不足的核心问题。现有方法过度依赖伪造数据，容易学习数据集特定的伪影，而非自然语音的可迁移特征。为此，作者