Method for power standardization speech recognition and dynamic matching of safety procedure ticket based on multiple models

By using a multi-model approach to identify speech signals in power field, semantic correction is performed using target acoustic models and language models, and fuzzy matching is performed using a safety ticket standard library. This solves the problems of accuracy and consistency in speech recognition in power field and enables efficient generation of standardized operation instructions.

CN122177111APending Publication Date: 2026-06-09GUANGZHOU POWER SUPPLY BUREAU GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU POWER SUPPLY BUREAU GUANGDONG POWER GRID CO LTD
Filing Date
2026-02-02
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

In power field operations, the recording and transcription of voice commands suffer from accuracy and stability issues. Existing technologies struggle to maintain high accuracy in voice recognition and standardized command conversion in complex environments.

Method used

A multi-model-based approach is adopted, which identifies spectrogram feature sequences through a target acoustic model, performs semantic correction and standardization processing by combining a language model, and finally uses the Ancuo Ticket Standard Library for fuzzy matching to generate accurate standardized operation instructions.

Benefits of technology

It improves the accuracy of speech recognition, eliminates errors in the manual confirmation process, achieves precise conversion from speech to standardized operations, and enhances the accuracy and consistency of operations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122177111A_ABST
    Figure CN122177111A_ABST
Patent Text Reader

Abstract

The application relates to a multi-model-based power standardized speech recognition and dynamic matching method for safety measures. The method comprises the following steps: inputting a spectrogram feature sequence corresponding to a single-channel speech signal of a power field into a target acoustic model constructed in advance to perform recognition processing, so as to obtain a preliminary text sequence; performing semantic correction and standardized processing on the preliminary text sequence by using a language model constructed in advance, so as to obtain a recognized text sequence; and performing fuzzy matching processing on the recognized text sequence by using a candidate standard sentence in a safety measure ticket standard library constructed in advance, so as to obtain a target matching result; wherein the target matching result is used to guide field operation of the power field. The method can improve speech recognition accuracy.
Need to check novelty before this filing date? Find Prior Art