首页> 外国专利> LEARNING METHOD AND TESTING METHOD FOR FIGURING OUT AND CLASSIFYING MUSICAL INSTRUMENT USED IN CERTAIN AUDIO AND LEARNING DEVICE AND TESTING DEVICE USING THE SAME

LEARNING METHOD AND TESTING METHOD FOR FIGURING OUT AND CLASSIFYING MUSICAL INSTRUMENT USED IN CERTAIN AUDIO AND LEARNING DEVICE AND TESTING DEVICE USING THE SAME

机译:学习方法和测试方法,用于确定某些音频和学习设备和测试设备中使用的乐器

摘要

According to the present invention, as a learning method for determining and classifying the tone of an instrument used for an arbitrary audio, (a) when an instrument audio corresponding to a specific instrument is input, the learning device analyzes the instrument audio to obtain a reference time. -The reference time is a time corresponding to each of a plurality of predetermined viewpoints among the total playing time of the instrument audio-and obtaining information on the sound intensity corresponding to each of the reference time as sound intensity trajectory data for each frequency; (b) the learning apparatus, (i) generating matrix data for each frequency with time and sound intensity as variables for each frequency with reference to the sound intensity trajectory data for each frequency, and (ii) the matrix for each frequency Performing a process of generating correction matrix data for each frequency by converting at least some of the element values included in the matrix data for each frequency by performing a predetermined operation on at least some of the element values included in data; And (c) performing, by the learning apparatus, inputting the correction matrix data for each frequency into a learning model as training data to perform training, and adjusting a plurality of parameters included in the learning model. Including, a learning method for determining and classifying musical instrument tone is provided.
机译:根据本发明,作为用于确定和分类用于任意音频,(a)的仪器的音调的学习方法,当输入对应于特定仪器的仪器音频时,学习设备分析仪器音频以获得a参考时间。 - 参考时间是对仪器音频的总播放时间中的多个预定视点中的每一个对应的时间,并且获得对应于每个频率的声音强度轨迹数据的每个参考时间的声强的信息; (b)学习装置,(i)在每个频率的每个频率的每个频率为每个频率的变量生成每个频率的矩阵数据,并且(ii)执行过程的每个频率的矩阵通过在包括在数据中的至少一些元素值上执行预定操作,通过对每个频率转换包括在矩阵数据中的至少一些元素值来为每个频率生成校正矩阵数据; (c)通过学习设备执行将每个频率的校正矩阵数据输入到学习模型中作为训练数据来执行训练,并调整包括在学习模型中的多个参数。包括提供用于确定和分类乐器音调的学习方法。

著录项

  • 公开/公告号KR102241436B1

    专利类型

  • 公开/公告日2021-04-19

    原文格式PDF

  • 申请/专利权人 주식회사 시스원;

    申请/专利号KR20190063493

  • 发明设计人 이진우;심은아;최한용;김상훈;

    申请日2019-05-29

  • 分类号G06N3/08;

  • 国家 KR

  • 入库时间 2022-08-24 18:26:40

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

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