Speaker Verification System for Emotional Voice
The project aims on the development of a speaker verification system that can handle emotional speech input.
Project GitHub: Not Ready.
The project is conducted by my 3rd year student, Prokhorova Maria, at HSE (Coursework) in 24/25 academic year.
Automatic Speaker Verification systems distinguish between the owner’s and another real person’s speech. However, existing solutions are mostly trained on datasets with neutral voice. This means that emotional speech, such as crying or screaming, will be out-of-domain for these systems and lead to poor performance. The project goal is to explore and improve the quality of ASV systems under emotional voice conditions.