A voice interaction method and system for virtual reality training platforms

By performing frame-segmentation and windowing processing on the speech signal and optimizing the weights of the MFCC coefficients, the speech recognition accuracy problem of the traditional MFCC algorithm under noise interference is solved, and the speech interaction effect of the virtual reality training platform is improved.

CN120279908BActive Publication Date: 2026-03-13UHV CO OF STATE GRID HEILONGJIANG ELECTRIC POWER CO LTD +1
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-04-18
Publication Date
2026-03-13

AI Technical Summary

Technical Problem

Traditional MFCC algorithms are subject to background noise interference during speech recognition, resulting in low speech recognition accuracy and affecting the effectiveness of voice interaction.

Method used

By performing frame-by-frame windowing on the speech signal, the stationarity index and semantic richness are obtained. Combined with the resonance noise level of the formants, the weighting factor of the MFCC coefficients is optimized using the particle swarm optimization algorithm, and weighted processing is performed to improve the speech recognition accuracy.

Benefits of technology

This reduces false positives due to noise caused by semantic features, improves the accuracy of speech recognition, and thus enhances the accuracy of voice interaction on the virtual reality training platform.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120279908B_ABST
    Figure CN120279908B_ABST
Patent Text Reader

Abstract

This application relates to the field of voice interaction technology, specifically to a voice interaction method and system for a virtual reality training platform. The method includes: acquiring a user-input voice signal; performing frame-by-frame windowing processing on the voice signal; acquiring the stationarity index of each frame of the voice signal; acquiring the pause intervals of the entire voice signal; acquiring the semantic richness of each frame of the voice signal; acquiring the optimized stationarity index of each frame of the voice signal; acquiring formants in the spectrogram of each frame of the voice signal; acquiring the resonance noise level of each formant; acquiring multiple MFCC coefficients of each frame of the voice signal; and acquiring weighting factors for each MFCC coefficient of a single frame of the voice signal, used to weight each MFCC coefficient, thereby completing the voice interaction of the virtual reality training platform. This application improves the accuracy of speech recognition by adaptively weighting each MFCC coefficient, thus improving the voice interaction accuracy of the virtual reality training platform.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Human voice highlighting processing method and device in audio

    CN104916288A

  • Intelligent office voice control method and system based on voice recognition

    CN117995178A