Latest Articles for Speech Technology

New methods improve speech recognition for whispered communication.

StyleSpeech advances TTS systems by capturing natural speech nuances.

EmoAttack leverages emotional voice conversion to exploit vulnerabilities in speech systems.

A new method improves converting whispered speech to normal speech using advanced techniques.

VoxInstruct combines content and style for more natural speech generation.

A novel method improves voice recognition accuracy across multiple languages.

Exploring a new approach to improving speech quality using time-context windowing.

New methods improve the quality of speech synthesis in TTS systems.

SelectTTS simplifies speech generation for unseen speakers with effective frame selection.

A new method improves speech model performance across various tasks.

A new method improves keyword spotting accuracy using unlabeled audio data.

Research shows speech analysis can aid in early detection of Mild Cognitive Impairment.

Researchers create LibriheavyMix to improve speech recognition in noisy environments.

A new benchmark aids in assessing speech tokenizers for better performance.

A new method leverages speech data to improve autism assessments.

Discover how DDSP improves speech synthesis efficiency and quality.

SpeechLLMs show promise but struggle with speaker identification in conversations.

This article discusses efficient training methods for speech models using self-supervised learning.

A new dataset enhances multilingual speech technology in India.

ParaEVITS improves emotional expression in TTS through natural language guidance.

Efforts to improve speech technology for the under-resourced Faetar language.

A new model combines speech recognition and entity recognition for better results.

A project aims to improve speech technology for those with communication challenges.

A new system enhances accent accuracy in TTS for better communication.

An easy-to-use tool for fine-tuning speech models without complex code.

A new method improving speech recognition while ensuring data privacy.

A new method for generating accented speech using text transliteration.

Wave-U-Mamba enhances low-quality speech recordings for clearer communication.

A new system predicts naturalness scores for synthetic speech using innovative methods.

Exploring the GenSEC challenge to improve speech transcription accuracy.

A new method assesses self-supervised speech models using rank measurement.

MCMamba model improves speech quality in noisy environments using spatial and spectral information.

A new framework enhances speech recognition by modeling sound relationships effectively.

A new approach enhances the interpretability of spoof speech detection.

A model improves speech tasks in multilingual settings, addressing code-switching challenges.

EVA combines audio and visual signals for better speech recognition accuracy.

A new method improves speech interactions by integrating recognition and response processes.

Research evaluates connections between speech and language models for improved recognition and translation.

Learn how to effectively train speech models with fewer labeled resources.

An analysis of gender terminology in speech technology and its societal implications.