BUBBLE DIVIDING METHOD FOR GENERATING COMPACT ACOUSTIC MODEL

摘要

PROBLEM TO BE SOLVED: To provide improved technology for constructing compact acoustic models for use in a large-vocabulary continuous speech recognition system. SOLUTION: The method for structuring the compact acoustic models includes: partitioning speech data from a plurality of training speakers according to at least one speech related criteria (i.e., vocal tract length); grouping together the partitioned speech from training speakers having a similar speech characteristic; and training an acoustic bubble model for each group using the speech data with the group. COPYRIGHT: (C)2005,JPO&NCIPI