一种用于调度电话的声纹识别处理方法
By decoupling voiceprint features using deep neural networks and orthogonal projection operators of channel subspace basis vector groups, the nonlinear distortion problem introduced by low bit rate encoding and decoding in dispatch telephones is solved, achieving stable and efficient identity recognition under multi-protocol channels.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INFORMATION & COMM CO OF STATE GRID SHAANXI ELECTRIC POWER CO LTD
- Filing Date
- 2026-05-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to effectively decouple the nonlinear distortions introduced by low bit-rate encoding and decoding in dispatching telephones, leading to irreversible collapse of voiceprint features in the manifold space and resulting in decreased recognition performance. In particular, it is difficult to maintain feature stability and purity under multi-protocol channels and short voice conditions.
A deep neural network model is used in conjunction with spectral entropy and channel subspace basis vectors. Orthogonal projection operators are used to decouple voiceprint features, eliminate channel distortion, and endpoint detection logic is used to filter out non-human voice segments. A dual orthogonal feature cleaning mechanism is constructed to ensure feature purity.
Maintaining the stability of voiceprint features under multi-protocol and extreme channel conditions ensures that recognition performance does not degrade, achieving compactness and accuracy in cross-channel identity recognition, and adapting to full-spectrum interference in complex scheduling scenarios.
Smart Images
Figure CN122177124B_ABST