| 摘要 |
<p>PROBLEM TO BE SOLVED: To provide more rapid talker identification processing technology. SOLUTION: The system for collating data includes a model creation part 5 which creates model data for collation from a series of LPC(linear predictive coding) cepstrum coefficients of inputted voice, a registered model storage part 11 which stores a plurality of registered model data which are the source of collation, an inter-model collation part 9 which performs collation processing by using the model data for collation and the registered model data, and a collated result deciding part 13 which specifies a registered model corresponding to top one or more registered model data by using the collated result between the model data for collation and the registered model data from the inter-model collation part 9. By performing multistage clustering of the registered models thereby reducing the number of objects to be collated or by performing more detailed collation processing using the result of collation between the models, identification exactness may be improved.</p> |
