A GMM-UBM Based Multi-speaker Re-segmentation and Re-clustering Algorithm

Aiming at the shortcomings of traditional speaker segmentation and clustering methods, this paper proposes a multilevel speaker re-segmentation and re-clustering algorithm based on GMM-UBM. The algorithm is based on the method of statistical modeling in the field of speaker recognition, and makes full use of the speaker information after segmenting and clustering in traditional methods to re-segment and re-cluster speech files, which improves the performance of the system effectively.