Johannes Traa—Artificial Intelligence

“Phase Difference Models and Nonnegative Matrix Factorization for (Audio) Source Separation”

Source separation is a significant problem in the audio realm. Today’s hand-held devices (e.g., cell phones) increasingly have multiple microphones that can be used to improve the quality of the audio extracted from the soundscape around the device. Spatial information is very useful in these contexts and it can be combined with single-channel source modeling techniques like Nonnegative Matrix Factorization (NMF). I will discuss some of my work in this area.