Audio denoising method and related apparatus, device, and storage medium
By combining prior estimation, diffusion mechanisms, and corrected spectrograms, a new audio denoising technology has been developed, which solves the problems in the existing audio denoising technology field, particularly in audio denoising methods, related devices, equipment, and storage media.
Patent Information
- Application Number
- CN202610433529.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-02
- Publication Date
- 2026-07-17
AI Technical Summary
Existing audio denoising techniques struggle to balance high fidelity and denoising capability while reducing the number of inference steps. In particular, prediction-based methods are prone to over-smoothing or distortion, while generative methods suffer from extremely high inference burden and latency.
A combined approach of prior estimation, diffusion mechanism, and spectrogram correction is adopted. Prior estimation is used to obtain a rough clean audio spectrogram, diffusion mechanism is used for noise reduction optimization, and possible lost information is supplemented in the spectrogram correction stage, thereby reducing the number of inference steps and improving the noise reduction quality.
While reducing the number of inference steps, it improves the fidelity and noise reduction capability of audio denoising, thereby enhancing denoising efficiency and quality.
Smart Images
Figure CN122417049A_ABST