Multilingual Speech Emotion Recognition (SER)
Overview
This project focuses on developing a universal Speech Emotion Recognition (SER) model capable of detecting emotions across multiple languages. The model leverages cutting-edge techniques such as Transformers, Convolutional Neural Networks (CNNs), and other state-of-the-art deep learning architectures to ensure accurate and culturally inclusive emotion recognition.
EmoTa: Tamil SER Dataset
As part of this project, EmoTa was developed and released to address the lack of resources for Dravidian languages in the SER domain. EmoTa is a Tamil SER dataset specifically designed to capture the nuances of Sri Lankan Tamil dialects, enhancing the model's ability to understand and process Tamil speech emotions effectively.
Objectives
- Build a universal SER model for 15+ languages, including Tamil.
- Enhance emotion recognition accuracy across diverse linguistic and cultural contexts.
- Provide high-quality datasets to support the development of multilingual SER systems.
Resources
- EmoTa Dataset Access: Dataset Link
- EmoTa Repository: GitHub Link