Embedding Techniques of Audio Segments Emphasizing Meaningful Semantic Characteristics

This project will develop artificial intelligence techniques to identify characteristics in audio samples that relate to real-world descriptions of sounds. When humans listen to various samples of sounds of a same category, they intuitively affix qualitative words to these sounds. The objective of the research is to find ways to let an AI algorithm find out these characteristics, and rank sounds based on them. For example, given a database of voice sound samples, the algorithm could automatically rank them in order of “kindness” in the voice, or “shyness”. The developed algorithms will help Ubisoft be more efficient and creative when designing the sound aspect of video games.

Faculty Supervisor:

Ghyslain Gagnon

Student:

Partner:

Ubisoft Toronto

Discipline:

Computer science

Sector:

Information and cultural industries; Manufacturing

University:

École de technologie supérieure

Program: