首页> 中文期刊>中国科技论文 >基于指纹权重的音频模板检索方法

基于指纹权重的音频模板检索方法

     

摘要

Aiming at the problem that the audio template retrieval method is not robust enough in the environment of noise and spectrum loss, a retrieval method based on the frame energy difference is proposed based on the traditional Philips retrieval method, in which the ignored energy information in the traditional method is used to effectively improve the robustness of the audio template matching method, by setting different matching weights on the DNA of different audio frames.The experimental results show that the accuracy of the audio search and the recall rate are improved under different lengths of audio templates on noise and spectrum missing data sets.For example, in the 2-second application environment, the precision rate is improved by 16%.%针对音频模板检索方法在噪音和频谱缺失等环境下鲁棒性不够强的问题,在原有Philips检索方法的基础上,提出了一种以帧能量差作为权重的检索方法,即充分利用原方法忽略的能量信息,通过对不同音频帧的DNA设置不同的匹配权重,有效提升了音频模板匹配方法的鲁棒性.实验结果表明,在受噪音影响和频谱缺失的数据集上,对于不同长度的音频模板,音频检索的查准率和查全率均有不同程度的提升,其中在2s的短模板上,查准率可比基线提高16%.

著录项

相似文献

  • 中文文献
  • 外文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号