Content:

The “Speaker Recognition: Principals and Practicalities” course presents the most important concepts in speaker recognition, such as feature extraction, classification and decision. Moreover, the course approaches performance evaluation metrics and score calibration methods in order to highlight the importance of post processing of recognition scores in real-world applications. Finally, the challenges such as speaker variability, channel variability, ambient noise and utterance duration are addressed and recent techniques to compensate for these undesirable effects are reviewed.

Objective

The course comprises of lectures and laboratory work. The laboratory work aims to make the student familiar with speaker recognition in practice. The students will develop a full front-end i-vector based speaker recognition in MATLAB using the provided toolbox. I-vector approach is the state-of-the-art method in speaker recognition. The course includes 4 lecture sessions and 4 lab sessions. The course runs in 4 consecutive days in week 7 having a morning lecture followed by afternoon lab session. Each session is going to be 2 times 45 minute with a 15 minute break.

The final evaluation is based on oral exam and the progress of student in implementing the full speaker recognition front-end. Interested students will have the possibility to do a course project within a week (week 8).

Exam

The oral exam would be on 18-19.2.