Skip to content

itmo-mbss-lab/sr_labs_book

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ITMO Speaker Recognition Course

Authors: Volokhov V., Lavrentyeva G., Novoselov S., Matveev Y.

Description: the project is related to the development of labs for the ITMO Speaker Recognition Course.

Keywords: voice biometrics, speaker recognition, speaker verification, speaker identification, acoustic features, speech activity detector, machine learning, speaker embedding extractor, deep neural network, decision theory, domain adaptation and calibration.

Datasets: the main databases for performing of labs is VoxCeleb corpus.

Fugure 1

Content: the repository contains materials (now only in russian language) for self-performing five labs. The titles of the labs are listed below.

  • Lab work 1. Informative features of speech signals: feature extraction (link).
  • Lab work 2. Voice activity detector training (link).
  • Lab work 3. Creating and comparing speaker models (link).
  • Lab work 4. Decision criteria and quality metrics (link).
  • Lab work 5. Adaptation and calibration of speaker recognition system (link).

Some ideas for creating of labs were borrowed here (training of voice activity detector model), here (training and testing of speaker embedding extractor) and here (training of calibration model for voice biometrics system).

Presentations of lectures on the course are presented below.

  • Lecture 1. Introduction to voice biometrics (link).
  • Lecture 2. Preprocessing of speech signals. Part 1 (link).
  • Lecture 3. Preprocessing of speech signals. Part 2 (link 1, link 2).
  • Lecture 4. Classical methods of constructing speaker models (link).
  • Lecture 5. Modern methods of constructing speaker models (link).
  • Lecture 6. Comparison of speaker models (link).
  • Lecture 7. Decision criteria. Quality assessment of biometric systems (link).
  • Lecture 8. Domain adaptation. Calibration of voice biometric systems (link).
  • Lecture 9. Where to go next? (link).
  • Lecture 10. Speaker diarization (link).
  • Lecture 11. Training neural network speaker models from transformer pretrains (link).

Video lectures of the course on the course are presented below.

  • Lesson 1. Review of lecture 1 (link).
  • Lesson 2. Review of lecture 1 (link).
  • Lesson 3. Review of lecture 2 and lab work 1 (link).
  • Lesson 4. Review of lecture 3 (link).
  • Lesson 5. Review of lecture 4 and lab work 2 (link).
  • Lesson 6. Review of lecture 4 (link).
  • Lesson 7. Review of lecture 5 (link).
  • Lesson 8. Review of lecture 5 (link).
  • Lesson 9. Review of lecture 6, 7 and lab work 3 (link).
  • Lesson 10. Review of lecture 7 and lab work 4 (link).
  • Lesson 11. Review of lecture 8 (link).
  • Lesson 12. Review of lecture 8 and lab work 5 (link).
  • Lesson 13. Review of lecture 9 (link).
  • Lesson 14. Review of lecture 10 (link).
  • Lesson 15. Review of lecture 11 (link 1, link 2, link 3).

A published version of these labs (now only in russian language) can be found here. Publication date: 05/24/2022.

A latest updated version of these labs (now only in russian language) can be found here. Publication date: 05/24/2022.