语音端点检测方法及相关装置、设备和介质
By employing a four-class state machine and refining noise types in speech endpoint detection, the problem of semantic fragmentation in speech endpoint detection is solved, thereby improving the semantic coherence of speech detection and the accuracy of downstream tasks.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ANHUI IFLYTEK UNIVERSAL LANGUAGE TECH CO LTD
- Filing Date
- 2025-10-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing speech endpoint detection technologies lack the ability to model the overall semantic level, leading to misjudgments of sentence termination during short pauses, causing semantic fragmentation and affecting the execution of downstream tasks.
A four-class state machine based on streaming audio is used for continuous monitoring. By refining the noise type into first noise, middle noise and last noise, and combining the energy value and probability value of the audio frame, the four-class state machine is used for continuous monitoring to determine the semantic position of the speech frame.
It improves the semantic coherence after speech endpoint detection, reduces incomplete sentence segmentation, and improves the accuracy of downstream tasks.
Smart Images
Figure CN121306102B_ABST