Spoken Language Technology. Speaker Recognition

Course in BIOSINF Master programme (1st year)

Teachers

Teachers: Prof. Corneliu Burileanu
Teaching Assistant: Assoc.Prof. Horia Cucu

Course Description

The “Spoken Language Technology. Speaker Recognition” course presents the most important concepts regarding speech analysis (production and modeling, perception, features extraction, etc.) and speech synthesis. Moreover, the course approaches speech analysis and synthesis from a hardware point of view, discussing speech signal processors. Finally, this course highlights the most important concepts in Speaker Recognition, such as speaker modeling, classification and decision.

The laboratory aims to make the student familiar with speech signal properties. It starts with the speech analysis in time and frequency and it continues with the configuration of several feature extraction methods. Several speech processing techniques are approached: pitch estimation, speaker recognition with Dynamic Time Warping and Gaussian Mixture Models.

Contents

Course

  1. The importance of speech analysis and synthesis systems
  2. Speech recognition strategy
  3. Acoustic and phonetic processor structures
  4. Speech signal analysis techniques
  5. Speech recognition techniques
  6. Artificial Neural Networks – ANN
  7. Speaker recognition

Laboratory

  1. Speech processing in time domain and frequency domain
  2. Speech features extraction methods
  3. Pitch estimation
  4. Speaker recognition with Dynamic Time Warping (DTW)
  5. Speaker recognition with Gaussian Mixture Models (GMMs)

Download

The course slides and the laboratory papers are available on Moodle.

Grading

Laboratory (Semester project + oral evaluation): 50%
Course final exam (oral evaluation): 50%