一种声纹识别方法、装置、终端及存储介质
By combining voiceprint models and CDS similarity algorithms, the problems of low accuracy and slow speed in existing voiceprint recognition technologies are solved, achieving more efficient voiceprint recognition and stronger defense against impersonation attacks.
CN116665680BActive Publication Date: 2026-07-17WONLY SECURITY & PROTECTION TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WONLY SECURITY & PROTECTION TECH CO LTD
- Filing Date
- 2023-06-27
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Current voiceprint recognition technologies have low accuracy and slow speed, and cannot effectively defend against impersonation attacks.
Method used
The voiceprint model includes a voiceprint background sub-model, a voiceprint classification sub-model, and a voiceprint recognition sub-model. Wavelet transform is used to extract the wavelet entropy of speech features, and the model is trained by a convolutional neural network. The CDS similarity algorithm is then used for voiceprint recognition.
Benefits of technology
It improves the accuracy and speed of voiceprint recognition and enhances the defense against impersonation attacks.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116665680B_ABST
Abstract
本发明公开了一种声纹识别方法、装置、终端及存储介质，包括：获取测试语音和训练语音，测试语音和训练语音均包括多个语音特征；将测试语音和训练语音分别输入到训练好的声纹模型中，获得测试语音中多个语音特征分别对应预设语音特征的第一后验概率矩阵，及训练语音中多个语音特征分别对应预设语音特征的第二后验概率矩阵；利用CDS相似度算法，对第一后验概率矩阵及第二后验概率矩阵进行相似度比较，得到所述测试语音的声纹识别结果。通过本发明通过分别计算测试语音和训练语音的后验概率矩阵，利用CDS相似度算法，对两个矩阵进行相似度比较，得到测试语音的声纹识别结果。通过本发明提供的方法提高了运算的速度及声纹识别的准确率。
Need to check novelty before this filing date? Find Prior Art