音频增强方法、装置、计算机设备和存储介质

By combining multi-level wavelet decomposition and reconstruction with a feature extraction model to optimize audio features, the problem of information loss in traditional audio enhancement methods is solved, and higher quality audio enhancement is achieved.

CN116564324BActive Publication Date: 2026-07-17INDUSTRIAL AND COMMERCIAL BANK OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
INDUSTRIAL AND COMMERCIAL BANK OF CHINA
Filing Date
2023-05-16
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Traditional audio enhancement methods, which rely on filtering, tend to miss crucial audio information, resulting in poor enhancement effects.

Method used

Multi-level wavelet decomposition is used to obtain frequency band features and wavelet coefficients. The correlation between adjacent frequency bands is then combined for reconstruction. Subsequently, the audio features are optimized through feature extraction and audio reconstruction models to finally obtain the target enhanced audio.

Benefits of technology

It improves audio enhancement, avoids missing key information, and enhances audio quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116564324B_ABST
    Figure CN116564324B_ABST
Patent Text Reader

Abstract

本申请涉及一种音频增强方法、装置、计算机设备、存储介质和计算机程序产品，可用于音频处理技术领域，也可用于金融科技领域或其他相关领域。所述方法包括：对金融系统的待处理音频进行多级小波分解处理，得到待处理音频的多个频带和每个频带对应的小波系数；获取每个频带对应的频带特征，根据每个频带的相邻频带对应的小波系数，对每个频带对应的小波系数进行重构处理，得到每个频带对应的重构小波系数；对每个频带对应的频带特征和重构小波系数进行小波重构处理，得到待处理音频的初始增强音频；根据待处理音频和初始增强音频的目标音频特征，进行音频重构处理，得到待处理音频的目标增强音频。采用本方法，能够提高音频增强效果。
Need to check novelty before this filing date? Find Prior Art