首页> 外国专利> Sound generator separable study manner, device and program, sound generator separable manner, device, program and record media

Sound generator separable study manner, device and program, sound generator separable manner, device, program and record media

机译:声音发生器可分离的学习方式,装置和程序,声音发生器可分离的方法,装置,程序和记录媒体

摘要

PPROBLEM TO BE SOLVED: To play back a target sound of a high S/N ratio by eliminating noise from the target sound collected by one microphone normally. PSOLUTION: The sound source separation learning apparatus uses signals collected by two microphones in advance to acquire first learning data constituted of weight values by bands for noise elimination that becomes a model case, and a class of an acoustic feature amount with which a weight value by bands most approximate to the first learning data, as second learning data. The sound source separation apparatus calculates a weight value by bands for noise elimination using the acoustic feature amount of the class designated by the second learning data for a signal collected by a single microphone, multiplies each of band dividing signals dividing the signal collected by the single microphone into bands by the weight value by bands, and eliminates a noise component. PCOPYRIGHT: (C)2006,JPO&NCIPI
机译:

要解决的问题:通过消除一个麦克风正常收集的目标声音中的噪声,播放高信噪比的目标声音。

解决方案:声源分离学习设备使用预先由两个麦克风收集的信号来获取第一学习数据,该第一学习数据由作为模型实例的用于消除噪声的频带的权重值以及以最接近作为第一学习数据的第一学习数据的频带的权重值,作为第二学习数据。声源分离设备使用由第二学习数据指定的针对单个麦克风收集的信号的类别的声学特征量,按频带计算权重值,以消除噪声,将除以单个麦克风收集的信号的每个频带划分信号相乘。麦克风按权重值分成频段,并消除了噪声成分。

版权:(C)2006,JPO&NCIPI

著录项

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