OpenAudioAI
The goal of this project is to eventually create an audiobook system that can autodetect who the speakers is and change its voice settings accordingly. To do that, our first step is to create a speaker segmentation system which can detect who the speaker is purely text-based. There is not very many papers about speaker detection that is purely text-based, so we created one here.
Author 1: Steffen Lim Email: slim13@students.kennesaw.edu Author 2: Sams Khan Email: skhan34@students.kennesaw.edu
https://docs.google.com/document/d/1fe0iCKSy-xPkjXN8e_PjmjxUOdGY7QqCyvKVCkFh2d8/edit?usp=sharing
To do: Dataset Preparation:
- [] 1) Split LSTM networks to context and main networks
-
[] 2) Implement concept from the main paper dynamic attention from sentence embedding to speaker embedding
Current model Idea:
