Matrix computations (3rd ed.)
Speaker recognition and speaker normalization by projection to speaker subspace
ICASSP '96 Proceedings of the Acoustics, Speech, and Signal Processing, 1996. on Conference Proceedings., 1996 IEEE International Conference - Volume 01
On the use of orthogonal GMM in speaker recognition
ICASSP '99 Proceedings of the Acoustics, Speech, and Signal Processing, 1999. on 1999 IEEE International Conference - Volume 02
Sparse principal component analysis via regularized low rank matrix approximation
Journal of Multivariate Analysis
Speaker Verification System Using Extended Sequential Probability Ratio Test in PDA
FMN '09 Proceedings of the 2nd International Workshop on Future Multimedia Networking
| Hi-index | 0.00 |
Prinicipal component anaysis (PCA) is one of the most general purpose feature extraction methods A variety of methods for PCA has been proposed Many conventional methods, however, require a larger amount of training data when the eigenvector matrix of each speaker is calculated This paper proposes a global eigenvector matrix based PCA for speaker recognition (SR) The proposed method uses training data from all speakers to calculate the covariance matrix and uses this matrix to find the global eigenvalue and eigenvector matrix to perform PCA During the training and testing of this method, the global PCA coefficients instead of the PCA coefficients of each speaker are used in performing PCA transformation Compared to the PCA and the conventional methods in the Gaussian mixture model (GMM)-based speaker identification (SI) and speaker verification (SV), the proposed method shows better performance while requiring less storage space.