Wednesday, September 20th at 10:30, Guolin Fang will defend his Ph.D. thesis proposal "Speech Synthesis Modelling using Epoch Features" in room M119. Teams link below.

Supervisor:

• Jón Guðnason, Department of Engineering, Reykjavík University
  Committee:
• Jacqueline Clare Mallet; Department of Computer Science, Reykjavik University
• Lauri Juvela; Department of Information and Communications Engineering, Aalto Universit

ABSTRACT: The central aim is a transformative advancement in End-to-End Text-to-Speech (TTS) systems, primarily through integrating classical speech signal processing techniques with modern machine learning. The essence of our innovation is the utilization of epoch characteristics, as opposed to the traditional method of using 25-millisecond windows. Traditional practices involve partitioning speech data into 25-millisecond windows, introducing a degree of distortion in speech synthesis. This distortion arises from the discontinuities in the speech signal due to the windowing process, thus impacting the quality of TTS outputs. By employing an epoch-based approach, we intend to capture each vocal cord vibration (epoch) in the speech signal, thereby avoiding distortion issues and improving the naturalness and resolution of speech synthesis.

Microsoft Teams meeting

Join on your computer, mobile app or room device

Click here to join the meeting

Meeting ID: 328 475 868 377
Passcode: LFDETJ

Download Teams | Join on the web