一种声学环境感知引导的自适应语音增强方法及装置
By employing an acoustic environment-aware adaptive speech enhancement method, which combines the Transformer architecture and the U-Net diffusion generation model, the problem of poor speech enhancement performance in multi-channel far-field environments is solved, achieving efficient speech enhancement and recognition under different acoustic environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH BEIJING
- Filing Date
- 2025-12-17
- Publication Date
- 2026-07-17
AI Technical Summary
In multi-channel far-field environments, existing speech enhancement methods suffer from the problem that beamformers are sensitive to sound source orientation errors and reverberation, leading to target speech distortion and poor speech enhancement effects.
An acoustic environment-aware adaptive speech enhancement method is adopted. This method performs time-frequency transformation on the multi-channel speech signals received by the microphone array, performs spatial sampling using a fixed beamforming filter, and combines an acoustic environment-aware network based on the Transformer architecture with a U-Net diffusion generation model to extract acoustic environment coding vectors and form joint features for speech enhancement.
It significantly improves the speech enhancement algorithm's performance in different acoustic environments, reduces speech distortion, and enhances speech intelligibility and speech recognition accuracy.
Smart Images
Figure CN121789701B_ABST