基于内容的音频检索具有较高的实用价值,将安静环境下训练的模型应用于实际有背景噪声的环境中。分析各种语音增强技术,并通过性能对比,选取谱减法作为系统前端抗噪技术,给出一个将音频增强和音频检索系统级联组成的抗噪声音频检索系统,并给出适合系统使用的谱减法的改进算法。
- Abstract:
- Content-based audio retrieval has high practical value.Under normal circumstances,the model parameter trained in the quiet environment is applied to environment which actually has background noise in retrieval system.This paper analyzes a variety of speech enhancement technology,and compares their performance,selects the spectral subtraction as noise-robust technology.A noise-robust audio retrieval system is given by linking audio retrieval system with an audio enhancement system,improved spectral subtraction algorithm suiting for retrieval system is provided.
声明:本文内容及配图由入驻作者撰写或者入驻合作网站授权转载。文章观点仅代表作者本人,不代表电子发烧友网立场。文章及其配图仅供工程师学习之用,如有内容侵权或者其他违规问题,请联系本站处理。 举报投诉