Talk in CSE: Group Delay Based Features and Feature Diversity for Speaker Recognition by R. Padmanabhan

Title: Group Delay Based Features and Feature Diversity for Speaker Recognition

Speaker: R. Padmanabhan
              School of Computing,University of Eastern Finland

Venue: IT Services Room

Date and time: Thursday 31 May, 11:30 pm

Overview of the Talk:

1. Introduction to speaker recognition.

2. Group delay based features for speech processing.

3. Feature diversity for speaker recognition.

Abstract:

1. Group delay based features for speech processing provide good performance.

2. Complementary information to magnitude based features like MFCC

3. Paradigm of feature-switching for speaker verification is introduced

4. Feature-switching improves EER by 2%

5. Feature-switching can be extended to any other verification task