The paper investigates how to improve self‑supervised learning (SSL) for Finnish speech by filtering broadcast audio based on automatic speech recognition (ASR) confidence rather than traditional voice activity detection or language identification. Using 2.1 million hours of raw broadcast audio, the authors demonstrate that ASR‑confidence filtering yields 3–7 % relative WER gains over LID filtering while training on a smaller dataset. They also conduct a controlled comparison of four SSL architectures—Transformer, Macaron‑Net, Conformer, and E‑Branchformer—at matched parameter budgets, showing that their models surpass previous Finnish SSL baselines, especially at Base scale and on out‑of‑domain data.
arXiv:2609.13045v1 Announce Type: new Abstract: Speech-to-speech translation (S2ST) has advanced significantly with speech LLMs, offering the potential for joint optimization and preserving non-lingu...
arXiv:2606. 08669v1 Announce Type: cross Abstract: Voice biometric systems face growing threats from spoofing attacks, yet the evaluation of detection models remains inconsistent across datasets.
arXiv:2606. 06743v1 Announce Type: cross Abstract: The popularity of neural audio codecs as speech tokenizers has surged with the advent of Multimodal Large Language Models.
arXiv:2606. 05678v1 Announce Type: cross Abstract: Automatic speech recognition (ASR) systems have become widely used for multilingual speech-to-text transcription.
arXiv:2609.13842v1 Announce Type: cross Abstract: Recent advances in speech synthesis and voice conversion have made deepfake speech increasingly realistic, making generalization to unseen spoofing a...
arXiv:2606. 30780v1 Announce Type: cross Abstract: Audio deepfakes are a growing challenge for the general public, as well as for journalists and fact-checkers.
The paper investigates whether the Voxtral audio‑language model can detect speech spoofing. It shows that without task‑specific adaptation, the model’s language‑model layers prioritize semantic content, making spoof‑discriminative acoustic cues less separable. By applying lightweight weight‑decomposed low‑rank adaptation (DoRA), the authors create Spooftral, which achieves an equal error rate of 4.25% on the ASVspoof5 evaluation set.
The study investigates which layers of the self‑supervised speech model WavLM carry the most spoof‑relevant information for spoof detection. Using the ASVspoof 5 database, the authors perform per‑layer linear probing, canonical correlation analysis (CCA) with acoustic features, and a grid search over layer subsets, finding that five specific layers outperform pooling all thirteen. CCA shows mid‑layers align with most acoustic features, while voice‑quality metrics correlate more with later layers, and visualizations suggest deeper layers encode general synthesis artifacts rather than attack‑specific signatures.
arXiv:2603. 29042v2 Announce Type: replace-cross Abstract: Phone recognition (PR) is a key enabler of multilingual and low-resource speech processing tasks, yet robust performance remains elusive.
arXiv:2609.38887v1 Announce Type: cross Abstract: Real-time voice conversion (VC) systems commonly rely on pretrained speaker embeddings from automatic speaker verification (ASV) models. While effect...
arXiv:2606. 29031v1 Announce Type: cross Abstract: In regulated domains such as banking and healthcare, where privacy constraints make real speech costly to collect and retain, synthetic speech from modern text-to-speech (TTS) is an appealing alternative for training automatic speech recognition (ASR) without exposing sensitive customer recordings.
arXiv:2508. 05149v2 Announce Type: replace-cross Abstract: Large language models (LLMs) have demonstrated potential in handling spoken inputs for high-resource languages, reaching state-of-the-art performance in various tasks.