A method for Acoustic Model Training, which includes a forced alignment of Speech Training Data, process data and obtain the estimated Scattering matrices of Scattering matrices, skew the intermediate class and Internal diagonalize matrices and in Intermediate Class Terna and Estimating eigenvectors, to obtain new functions using discriminative vectorTrain a New Acoustic Model, to Save the Model, method and System for Acoustic Model Training

摘要

A method for Acoustic Model Training, which includes a forced alignment of Speech Training Data, process data and obtain the estimated Scattering matrices of Scattering matrices, skew the intermediate class and Internal diagonalize matrices and in Intermediate Class Terna and Estimating eigenvectors, to obtain new functions using discriminative vectorTrain a New Acoustic Model, to Save the Model, method and System for Acoustic Model Training