METHOD FOR EXPRESSING ASSEMBLY OF ACOUSTIC UNITS IN COMPACT MODE AND CHAINING TEXT-SPEECH SYNTHESIZER SYSTEM

摘要

PURPOSE: To compactly express a set of acoustic section units for text-speech synthesis by using a spectrum encoding method based on an acoustic ergodic hidden Markov model(AEHMM). CONSTITUTION: The text-speech synthesizer system includes processors 101 to 104, etc., constructs an AEHMM and initializes the AEHMM by a vector quantizing clustering system having the same size as the AEHMM. Then the system trains the AEHMM by back-andforth algorithm and a Baum Welch reevaluation formula and constructs a set of sectional units to be used for speech synthesis. The sectional units are encoded by the labeling feature of the AEHMM and a set of sectional units to be connected like a chain for speech synthesis is decoded by an AEHMM decoder.