Speaker retrieval based on minimum distance in HCT

In hierarchical cellar tree (HCT), this paper proposes a novel and effective speaker retrieval method based on minimum distance, which can be much better used to calculate the distance of unequal length vectors. In speaker indexing stage, minimum distance is used to control cell mitosis and HCT growing. In speaker retrieval stage, minimum distance is used to retrive the most relevant speakers. Experimental results have demonstrated that the minimum distance can be much better than some common used distance formulas.