Latest Articles for Speech Technology

Researchers improve detection of machine-generated speech using phase information adjustments.

A new framework enhances the study of unsupervised speech recognition systems.

New model LinDiff improves speech synthesis speed and quality.

Researchers blend visual and sound features to improve speech for electrolarynx users.

This research highlights how LLMs enhance speech understanding in long videos.

A new method optimizes speech models for better performance with fewer resources.

EM-Network enhances sequence learning in speech and language processing tasks.

This study assesses various models for predicting synthesized speech quality.

This article discusses enhancing speech recognition using confidence-based ensemble methods.

GenerTTS enhances text-to-speech technology for cross-lingual applications.

A new model improves speech extraction from noisy backgrounds using deep learning.

A study on improving vocal sound reproduction through advanced synthesis techniques.

New methods aim to hide speaker identities while maintaining speech clarity.

A new method to improve speech quality using energy-efficient networks.

Researchers analyze how emotions are shared through speech using diverse data.

New methods improve the quality of synthesized speech using self-supervised learning.

Federated Learning improves speech recognition while keeping user data private.

A new method improves emotion detection from speech using audio only.

O-1 improves speech recognition by optimizing self-training methods.

Research highlights real-time detection methods for fake audio created by AI.

New pruning methods enhance zero-shot multi-speaker text-to-speech model performance.

New methods for selecting speech data minimize labeling while improving recognition accuracy.

A new method enhances speech quality ranking using listener preference scores.

A method to enhance ASR systems for users who stutter.

New single-step methods improve accuracy in formant tracking for speech sounds.

A new approach enhances the integration of speech with language models.

Examining how pretrained language models improve text-to-speech quality.

Microsoft's MuLanTTS offers natural and expressive French text-to-speech capabilities.

A project aims to improve French speech processing using self-supervised learning.

New methods improve how machines recognize speech rhythm and emotion.

This study improves ASR systems' ability to recognize children's speech.

VoxtLM combines speech recognition, synthesis, text generation, and continuation in one model.

Libriheavy offers 50,000 hours of spoken English to boost speech recognition technology.

AV2Wav enhances speech quality using audio and visual cues.

Core-set selection improves text-to-speech models by focusing on diverse data.

New method preserves emotional tone in voice conversion for better human-computer interaction.

Research reveals emotional speech impacts model performance in speech separation tasks.

Research combines self-supervised learning and new measurement techniques for improved speech inversion.

Researchers develop a new framework to enhance speech clarity for electrolaryngeal users.

A new method uses synthetic data to enhance ASR systems in unfamiliar areas.