语音质量检测方法和相关方法、装置、设备和存储介质
By extracting acoustic features of speech and referencing a dictionary, identifying phoneme probabilities and candidate pronunciation paths, and calculating pronunciation quality indicators, the problem of low accuracy in pronunciation quality detection during speech recognition is solved, achieving higher detection accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- IFLYTEK CO LTD
- Filing Date
- 2024-01-09
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, the accuracy of target speech pronunciation quality detection is low due to the influence of reference context and the relationship between the preceding and following text during the speech recognition process.
By extracting the acoustic features of the target speech, phoneme probabilities and text recognition are based on the acoustic features. Candidate pronunciation paths are obtained by combining the reference dictionary. The detection values of pronunciation quality indicators are calculated using the phoneme probabilities of audio frames and candidate pronunciation paths, providing reference information on the standard and actual pronunciation of the target speech.
It improves the accuracy of target speech pronunciation quality detection, especially in the absence of reference text, and can more accurately assess the pronunciation quality of speech.
Smart Images
Figure CN117912450B_ABST