| 摘要 |
PURPOSE:To obtain a voice recognition device which is hardly affected by the fluctuation in the intensity of phonatory deformation. CONSTITUTION:This phonatory deformation voice recognition device has a phonatory deformation voice recognition means 6 composed of a deformation degree memory 16 which stores deformation degree parameters, a multiphonatory deformation model forming means 17 which forms a phonatory deformation model with the phonatory deformation model stored in a phonatory deformation model memory as input according to the deformation degree parameters stored in the deformation degree memory 16, a spectrum deforming means 8 which subjects the phonatory deformation-less voice standard model stored in a phonatory deformation-less voice standard model memory 5 to spectrum deformation processing according to the phonatory deformation model which is the output of this multiphonatory deformation model forming means 17, a selection type likelihood calculating means 18 which executes likelihood calculation of an input voice characteristic vector time series and the phonatory deformation standard model which is the output of the spectrum deforming means 8, a collating mean 11 which executes collation processing by using the output of this selection type likelihood calculating means 18 and outputs the results of the recognition. |
