The Speaker Verification System Based on GMM Adaption Clustering and i-vector
Hong Zhao, Zhan Ling · 2020
The speaker verification system based on i-vector confirms the poor robustness and the discriminability of LDA is not strong. To solve this problem, the GMM adaptive clustering algorithm is used in this paper which is contributed to reduce the size of GMM, then the cluster center model is used to generate GMM supervector adopted to extract i-vector. Meanwhile, the MFD algorithm is improved which obtain the median value after removing the maximum and minimum value of i-vectors. Combined with GPLDA model, the simulation results show that the improved algorithm is an effective method.