Voice activity intelligent detection method and device, electronic equipment and storage medium
By identifying the communication start word and adjusting the VAD threshold and sliding time window length in speech activity detection, combined with Mel frequency cepstral coefficient feature matching, the problem of low detection accuracy in existing technologies is solved, and high-accuracy speech activity detection under dynamic environmental noise is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGZHOU HOTAPPS TECH LTD
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing speech activity detection methods cannot quickly adapt to dynamic changes in environmental noise levels, resulting in low detection accuracy.
By identifying the presence of a communication start word in the audio signal to be identified, the speech activity detection (VAD) threshold is determined based on the energy of the audio frame to be identified. The sliding time window length is adjusted in conjunction with the jitter rate. The Mel frequency cepstral coefficient feature is used to match the start word feature library, and the VAD threshold and window speech energy are dynamically adjusted to improve detection accuracy.
It enables voice activity detection that can quickly adapt to dynamic environmental noise, improves detection accuracy, reduces configuration complexity, and is suitable for resource-constrained embedded devices and real-time communication systems.
Smart Images
Figure CN122417064A_ABST