16 114 833 książek w 175 językach
Jednak się nie przyda? Nic nie szkodzi! U nas możesz zwrócić towar do 30 dni
Bon prezentowy to zawsze dobry pomysł. Obdarowany może za bon prezentowy wybrać cokolwiek z naszej oferty.
30 dni na zwrot towaru
Due to the growing need for security applications,§speaker recognition as the biometric task of§authenticating a claimant by voice has currently§become a focus of interest.§In this book we present new approaches to integrate§discriminative classifiers like Support Vector§Machines (SVMs) and Sparse Kernel Logistic Regression§(SKLR) into speaker recognition systems that are§traditionally based on generative classifiers like§Gaussian Mixture Models (GMMs).§In a first approach for limited training data the§discriminative classifiers are applied directly on§feature vectors from parameterized speech frames and§it is shown that both, SVM as well as SKLR outperform§traditional methods.§In the second approach a state-of-the-art speaker§recognition system for large amount of training data§is designed that combines Gaussian Mixture Models§with discriminative classifiers.§Furthermore, we investigate different feature§extraction methods for speaker recognition on large§amount of training data and it is shown that the§application of fusion schemes that combine these§subsystems yield a significant improvement of the§recognition performance in comparison to the§application of single subsystems.