A Novel Speaker Clustering Algorithm in Speaker Recognition System

Speaker clustering is involved in serial structure speaker identification system to reduce the algorithm delay and computational complexity. The speech is first classified into speaker group, and then searches the most likely one inside the group. Difference between Gaussian mixture models (GMMs) is widely applied in speaker classification. The paper proposes a novel measure based on pseudo-divergence, the ratio of inter-model dispersion to intra-model dispersion, to denote the difference between GMMs. And the measure is used to perform speaker clustering. Experiments indicate that the measurement works well to denote the difference of GMMs and has improved performance of speaker clustering