Top

Papers on speaker segmentation and polynomial matrix eigenvalue decomposition accepted for ICASSP’21

Aidan Hogg’s work [1] proposes to track the pitch of human voices in order to segment the onsets and endpoints of multiple competing speakers. Vincent Neo’s work [2] proposes to focus on speech in noise and reverberation by steering beams in the source direction and decomposing the resulting signals into a source and noise subspace.

[1] Hogg, Aidan, Evers, Christine and Naylor, Patrick A. (2021) Multichannel overlapping speaker segmentation using multiple hypothesis tracking of acoustic and spatial features. accepted for IEEE International Conference on Acoustics, Speech, & Signal Processing (ICASSP).

[2] Neo, Vincent W., Evers, Christine and Naylor, Patrick (2021) Polynomial matrix eigenvalue decomposition of spherical harmonics for speech enhancement.  accepted for IEEE International Conference on Acoustics, Speech, & Signal Processing (ICASSP).

No Comments

Post a Comment