Articulatory-feature based sequence kernel for high-level speaker verification

作者：Zhang Shi Xiong^*; Mak Man Wei

来源：7th International Conference on Machine Learning and Cybernetics, 2008-07-12 to 2008-07-15.

• speaker verification
• kernels
• articulatory features
• pronunciation models
• SVM

Research has shown that articulatory, feature-based phonetic-class pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However., the scoring method used in AFCPMs does not, explicitly use the discriminative information available in file training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the supervectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.

• 出版日期2008
• 单位香港理工大学