METHOD FOR COMPOSING CLASSIFICATION NEURAL NETWORK OF OPTIMUM SECTION AND AUTOMATIC LABELLING METHOD AND DEVICE USING CLASSIFICATION NEURAL NETWORK OF OPTIMUM SECTION

摘要

PPROBLEM TO BE SOLVED: To provide a method for composing a classification neural network of an optimum section and an automatic labelling method and device using classification neural network of the optimum section. PSOLUTION: The automatic labelling method using classification neural network of the optimum section comprises a step (a) that searches a neural network having a minimum error for each of L combinations of phoneme from K neural network sets created or updated in an initial stage, learns the K sets of the neural networks by the group of K combinations of phoneme searched from the same neural network and updates the weighted value, and composes K neural networks obtained at the time when the sum of the total error of K neural networks is converged with individual errors converged with the classification neural network sets of the optimum section; and a step (b) that corrects a boundary of phoneme of the primary label file using the phoneme combination group classification result and the classification neural network sets of the optimum section provided in the step (a), and creates a final label file reflecting the correction result. PCOPYRIGHT: (C)2004,JPO&NCIPI