Sriram Ganapathy

Integrating metadata enhances performance in speech tasks like language identification.

Combining foundational and specialized models boosts AI capabilities efficiently.

New methods combine audio and metadata for better language recognition.

Learn how dereverberation boosts speech recognition in noisy environments.

E-SHARC improves speaker identification in various audio environments.

This article presents a dual encoder system for effective speech representation learning.

New method improves ASR systems' handling of various accents through specialized codebooks.

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

A novel method combines meaning and sound for improved emotion detection in speech.

New methods improve understanding of AI model predictions.

Examining how our brains process sound and speech in different situations.