Presentation The objective of this course is to introduce students to advanced methods and techniques in the field of audio signal processing, with a strong applicative focus on the human voice which is at the center of human-computer interaction systems. This course aims in particular to combine the skills of students already acquired in acoustics, signal processing, and machine learning to solve complex tasks in the field of automatic speech processing and more generally for scene or sound environment analysis.
Fundamentals of acoustics: vibration, sound wave propagation, room acoustics
Sound signal analysis and synthesis: physical model of speech production, source/filter model of speech signal, and phase vocoder
Sound source localization: channel formation, binaural localization
Sound source separation: factorization into non-negative matrices
Automatic speech processing: spoken communication, general linguistics, text-to-speech synthesis
A large part of the course consists of a 16-hour project to be chosen from among those proposed, and the deepening of one of the aspects seen in the course.