一种基于神经网络的微量级嵌入式系统语音增强方法、装置、电子设备及存储介质
By using a neural network-based micro-level speech enhancement method in embedded systems, and combining filter banks and pre-trained neural networks, the problems of poor noise reduction effect of traditional algorithms and high computational cost of neural networks are solved, thus achieving efficient noise reduction in embedded systems.
CN120564739BActive Publication Date: 2026-07-17SHENZHEN JIAYZ PHOTO IND LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN JIAYZ PHOTO IND LTD
- Filing Date
- 2025-06-06
- Publication Date
- 2026-07-17
Smart Images
Figure CN120564739B_ABST
Abstract
本发明公开了一种基于神经网络的微量级嵌入式系统语音增强方法、装置、电子设备及存储介质。该方法包括获取待处理语音数据的功率谱数据，并为预设滤波器组计算滤波器组系数；将功率谱数据和滤波器组系数输入预训练神经网络，得到初始维纳滤波系数，预训练神经网络通过比率掩码方式和信号近似方式训练得到；对初始维纳滤波系数进行插值处理，得到与待处理语音数据频点数量相同的目标维纳滤波系数；通过目标维纳滤波系数处理待处理语音数据，得到降噪处理后语音数据。本发明的技术方案在减少降噪计算量的同时提升降噪效果。
Need to check novelty before this filing date? Find Prior Art