Excitation transformer voiceprint library construction method based on denoising reconstruction and two-stage classification

Abstract

The present application relates to the technical field of voiceprint recognition, and discloses a method for constructing a field current transformer voiceprint library based on denoising reconstruction and two-stage classification, which comprises the following steps: collecting voiceprint data of the field current transformer and performing discrete wavelet transform to obtain low-frequency main components; performing variational mode decomposition on the low-frequency main components to obtain intrinsic mode functions, and selecting intrinsic mode functions in the low-frequency band for reconstruction to restore the field current transformer body voiceprint features; obtaining working state data of the field current transformer and splicing the working state data with the field current transformer body voiceprint features to form fusion features, classifying the fusion features by using a voiceprint recognition model based on two-stage multi-level classification, and determining the voiceprint category; constructing a field current transformer voiceprint library according to the voiceprint category, collecting new voiceprint data, and performing similarity matching between the new voiceprint data and the field current transformer voiceprint library, and updating the field current transformer voiceprint library according to the matching result. The method improves the accuracy of field current transformer state monitoring and the intelligent level of voiceprint library construction.

Images