论文解读
Why Do You Say It Like That? A Phoneme-Level Framework for Explainable Speech Deepfake Detection
语音伪造检测 | 6.5/10
音频理解
论文解读
CARD: Cross-component Audio Representation Distillation for Encoder-Free Audio Captioning
音频字幕生成 | 6.3/10
论文解读
CaReCoS: A Spectrogram based Visual Benchmark for Cardiac, Respiratory and Cough Sounds
音频理解 | 6/10
论文解读
Probing Low-Level Acoustic Attribute Encoding in CLAP Audio Embeddings
音频理解 | 6.2/10
论文解读
Unified Audio Intelligence Without Regressing on Text Intelligence
音频交互 | 6.8/10
论文解读
AudioChat: Unified Audio Storytelling, Editing, and Understanding with Transfusion Forcing
音频生成 | 5.8/10
论文解读
AuTAgent: A Reinforcement Learning Framework for Tool-Augmented Audio Reasoning
音频理解 | 6.2/10
论文解读
MECAT: A Multi-Experts Constructed Benchmark for Fine-Grained Audio Understanding Tasks
音频理解 | 9.1/10
论文解读
MedMosaic: A Challenging Large Scale Benchmark of Diverse Medical Audio
音频理解 | 6.4/10
论文解读
Multimodal Fact-Level Attribution for Verifiable Reasoning
音频理解 | 6.4/10
论文解读
SPEAR: A Unified SSL Framework for Learning Speech and Audio Representations
音频理解 | 8.4/10
论文解读
Speech-Audio Compositional Attacks on Multimodal LLMs and Their Defense with SALMONN-Guard
音频理解 | 8.3/10
论文解读
Adaptive Perturbation Selection for Contrastive Audio Decoding
音频理解 | 5.3/10
论文解读
ORCA: Open-ended Response Correctness Assessment for Audio Question Answering
音频理解 | 7.9/10
论文解读
SpeakerLLM: A Speaker-Specialized Audio-LLM for Speaker Understanding and Verification Reasoning
说话人验证 | 7/10
论文解读
Audio-DeepThinker: Progressive Reasoning-Aware Reinforcement Learning for High-Quality Chain-of-Thought Emergence in Audio Language Models
这篇论文旨在解决大型音频语言模型（LALMs）缺乏显式、高质量推理能力的问题。现有方法要么受限于监督数据的质量，要么使用粗糙的奖励，导致生成的思维链形式良好但缺乏声学依据。作者提出了**Audio-DeepThinker**框架，其核心贡献有三：1）设计了一种**混合推理相似度奖励**，结合LLM评
论文解读
A Manual Bar-by-Bar Tempo Measurement Protocol for Polyphonic Chamber Music Recordings: Design, Validation, and Application to Beethoven's Piano and Cello Sonatas
本文旨在解决现有自动化节拍提取工具在分析历史复调室内乐录音（特别是贝多芬钢琴与大提琴奏鸣曲）时出现的系统性失败问题。作者与一名VLSI工程师合作，设计并验证了一套形式化的手动逐小节速度测量协议。该协议
论文解读
Beyond Transcription: Unified Audio Schema for Perception-Aware AudioLLMs
这篇论文旨在解决当前音频大语言模型（AudioLLMs）在细粒度声学感知任务上表现不佳的核心问题。作者指出，主流的以自动语音识别（ASR）为中心的训练范式，通过将音频映射到纯文本转录，系统性地丢弃了副
论文解读
Elastic Net Regularization and Gabor Dictionary for Classification of Heart Sound Signals using Deep Learning
本文旨在解决心音信号（PCG）的多分类问题，以辅助心血管疾病的自动诊断。核心贡献在于提出了一套结合**优化Gabor字典**和**弹性网络正则化**的特征提取框架，并与**CNN-LSTM深度学习网络
论文解读
Enhancing time-frequency resolution with optimal transport and barycentric fusion of multiple spectrogram
**核心问题**：短时傅里叶变换（STFT）生成的谱图受制于不确定性原理，无法同时获得优异的时间和频率分辨率。传统融合方法（如几何平均）要求输入谱图网格对齐，且性能有限。 **核心方法**：本文提出一