This paper investigates the correlation between dimensions of self-supervised speech features and speech characteristics, finding that content dimensions primarily capture intensity, formants, and voi…

This paper proposes an SSL-AutoEncoder (SSL-AE) approach for reducing feature dimensions in self-supervised learning models while maintaining dysarthric ASR performance.

The paper introduces a novel Clean-Referenced Feature-Vocoder Attack, a black-box adversarial attack that perturbs high-level SSL feature representations instead of raw audio waveforms, achieving supe…

HoliTok introduces a novel continuous holistic tokenization model that provides a unified, high-fidelity latent representation for simultaneously supporting both speech generation and speech understan…

The paper proposes SALMONN-2, an ALLM built on a unified SSL encoder, and presents a multi-layer feature fusion adapter to better exploit hierarchical SSL encoder representations. It also explores mul…

The paper proposes methods for generating global prosodic embeddings using auto-encoder models of pitch and energy, demonstrating competitive or superior performance under challenging conditions.

This paper investigates the relationship between language model perplexity and word error rate in modern automatic speech recognition systems, studying the impact of external language models, encoder…

PolySpeech-100 introduces a massive, multi-lingual benchmark covering 110 linguistic variants to rigorously test Speech-LLMs, demonstrating that open-source models struggle with low-resource languages…

The paper introduces SURE, a unified framework designed to standardize and improve the comparability and reproducibility of evaluations for advanced speech understanding models.

MOSS-Audio is a unified audio-language model designed for comprehensive understanding of speech, environmental sounds, and music, achieving strong performance across various audio-grounded tasks.

The paper proposes SpeechKV, a method to compress speech sequences inside large language models using a learned pooling, maintaining performance and delivering decoding speedup.

This paper investigates the use of various acoustic features for recognizing dysarthric speech using a Factorized Time Delay Neural Network (F-TDNN) model, achieving a relative improvement of 4.65% in…

This paper reviews various mechanisms for calculating word embeddings, investigates popular toolkits and matrices, and experiments with selected implementations.

This paper analyzes speech-text interleaved language models and reveals that they go through an implicit transcription phase in which spoken words become decodable as text in intermediate layers.

MindVoice is a neuro-to-speech framework that uses pretrained priors to disentangle and reconstruct intelligible speech from noisy, non-invasive neural signals, significantly outperforming existing me…

This paper proposes a segment-level representation learning framework for speech-based cognitive impairment detection using autoencoders and contrastive objectives.

This paper conditions a diffusion-based speech enhancement model on wav2vec 2.0 features using Feature-wise Linear Modulation (FiLM), achieving competitive performance on VoiceBank-DEMAND and LibriMix…

This paper proposes a novel layer-wise decision fusion method for fake audio detection using deep speech models, achieving the best cross-dataset performance.