Speech recognition result correction method and electronic device

By generating segmentation vectors and matching them with a text library, and combining historical recognition results with pre-stored text, the speech recognition results are optimized, solving the problem of inaccurate speech recognition caused by the diversity of user pronunciation, and improving the response accuracy of electronic devices and user experience.

CN122364449APending Publication Date: 2026-07-10HONOR DEVICE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HONOR DEVICE CO LTD
Filing Date
2025-01-09
Publication Date
2026-07-10

Smart Images

  • Figure CN122364449A_ABST
    Figure CN122364449A_ABST
Patent Text Reader

Abstract

This application relates to the field of terminal device technology, and provides a method for correcting speech recognition results and an electronic device. The method includes: obtaining a first text corresponding to a received speech command; obtaining a first segmentation vector based on the feature values ​​of the initials and finals corresponding to all characters in the first text; matching the first segmentation vector with a preset text library to obtain a text to be selected, and a first influence factor corresponding to the text to be selected; and determining the command text in the text to be selected based on the first segmentation vector and the first influence factor. Based on the technical solution provided by this application, after the electronic device recognizes the text corresponding to the speech command, it matches the segmentation vector of the text with the content in the text library to correct the recognition result through the preset text library, thereby reducing the problem of inaccurate response caused by speech recognition errors of the electronic device.
Need to check novelty before this filing date? Find Prior Art