Sound Analysis-Synthesis

Sound analysis encompasses methods that enable the extraction or automatic structuring of various types of information from the signal, such as the fundamental frequency or spectral evolutions determining the pitch and timbre of the perceived sound.

Information that is not strictly musical is also taken into account and is relevant to fields such as industrial acoustics, sound design, and multimedia. One notable example is the automatic indexing of sound recordings. The methods employed rely on signal processing, statistical analysis, machine learning and deep learning techniques, pattern recognition, as well as knowledge of auditory perception.

Sound transformation and synthesis techniques are initially developed to meet the needs of musicians in creating new sounds and new musical forms. This work also finds numerous applications in fields such as cinema, video games, navigation assistance, and more broadly virtual reality. Typical examples include computer-based singing synthesis and the creation of transformed and augmented voices and instruments. Research in analysis and synthesis relies on the design, on the one hand, of signal models that make it possible to describe and reproduce sound production using signal-processing methods and, on the other hand, deep learning models aimed at capturing correlations between sounds and the targeted characteristics. These models are implemented as software for computers (macOS, Windows, and Linux), specifically designed for professional or non-professional users: musicians, sound engineers, acousticians, and enthusiasts.