Greedy Kernel PCA를 이용한 화자식별 = Speaker Identification Using Greedy Kernel PCA

다국어 초록 (Multilingual Abstract)

In this research, we propose a speaker identification system using a kernel method which is expected to model the non-linearity of speech features well. We have been using principal component analysis (PCA) successfully, and extended to kernel PCA, which is used for many pattern recognition tasks such as face recognition. However, we cannot use kernel PCA for speaker identification directly because the storage required for the kernel matrix grows quadratically, and the computational cost grows linearly (computing eigenvector of $l{\times}l$ matrix) with the number of training vectors I. Therefore, we use greedy kernel PCA which can approximate kernel PCA with small representation error. In the experiments, we compare the accuracy of the greedy kernel PCA with the baseline Gaussian mixture models using MFCCs and PCA. As the results with limited enrollment data show, the greedy kernel PCA outperforms conventional methods.

참고문헌 (Reference)

1 J.P.Campbell,Jr., "Testing with the YOHO CD-ROM voice verification corpus" 341-344, 1995

2 D.A.Reynolds, "Robust text-independent speaker identification using Gaussian mixture speaker models" 3 (3): 72-83, 1995

3 R. O. Duda, "Pattern Classification" Wiley-Interscience Publication 2000

4 V. Franc, "Optimization Algorithms for Kernel Methods" Czech Technical University 2005

5 B.Scholkopf, "Nonlinear component analysis as a kernel eigenvalue problem" 10 (10): 1299-1319, 1998

6 B.Scholkopf, "Kernel principal component analysis" 583-588, 1997

7 J. Shawe-Taylor, "Kernel Methods for Pattern Analysis" Cambridge University Press 2004

8 T.-J.Chin, "Incremental kernel principal component analysis" 16 (16): 1662-1674, 2007

9 B.Scholkopf, "Fast approximation of support vector kernel expansions, and an interpretation of clustering as approximation in feature spaces" 124-132, 1998