Paper: Spherical Discriminant Analysis in Semi-supervised Speaker Clustering

ACL ID N09-2015
Title Spherical Discriminant Analysis in Semi-supervised Speaker Clustering
Venue Human Language Technologies
Session Short Paper
Year 2009
Authors

Semi-supervised speaker clustering refers to the use of our prior knowledge of speakers in general to assist the unsupervised speaker clustering process. In the form of an in- dependent training set, the prior knowledge helps us learn a speaker-discriminative fea- ture transformation, a universal speaker prior model, and a discriminative speaker subspace, or equivalently a speaker-discriminative dis- tance metric. The directional scattering pat- terns of Gaussian mixture model mean su- pervectors motivate us to perform discrimi- nant analysis on the unit hypersphere rather than in the Euclidean space, which leads to a novel dimensionality reduction technique called spherical discriminant analysis (SDA). Our experiment results show that in the SDA subspace, speaker clustering yields su- per...