METHOD AND DEVICE FOR RUN TIME ACOUSTIC UNIT SELECTION FOR LANGUAGE SYNTHESIS

摘要

PROBLEM TO BE SOLVED: To form a speech, which is naturally sound generated, by selecting one instance of each acoustic unit from many instances and connecting the selected instances. SOLUTION: When an input device 14 in a speech synthesis system receives text data, the inputted text data are transmitted to a synthesis element 15 which executes a speech synthesis. A speech synthesizer 36 converts a sound generating string to a diphone or the string corresponding to other acoustic unit, selects a best instance against each unit, adjusts the instance based on rhythmical parameters and connects one instance of each acoustic unit. The selection of an instance is conducted by enumerating thinkable sequences of the instances existed in the inputted language expression and selecting the instance which makes the spectrum distortion among all boundaries of adjacent instances of the sequence a minimum.