首页>
外国专利>
SOUND SOURCE ENHANCEMENT DEVICE, SOUND SOURCE ENHANCEMENT LEARNING DEVICE, SOUND SOURCE ENHANCEMENT METHOD AND PROGRAM
SOUND SOURCE ENHANCEMENT DEVICE, SOUND SOURCE ENHANCEMENT LEARNING DEVICE, SOUND SOURCE ENHANCEMENT METHOD AND PROGRAM
展开▼
机译：声源增强设备，声源增强学习设备，声源增强方法和程序
展开▼
页面导航
摘要
著录项
相似文献
摘要
PROBLEM TO BE SOLVED: To provide a sound source enhancement technique using a time frequency mask estimated by deep layer learning utilizing frequency conversion defined by a real number.;SOLUTION: A sound source enhancement processing section comprises: a frequency conversion section for generating an observation signal frequency conversion spectrum XtC from an observation signal χt by using frequency conversion defined by a real number; a time frequency mask estimation section for estimating a time frequency mask ^GtC from an acoustic feature amount ψt extracted from the observation signal frequency conversion spectrum XtC by a neural network using a parameter Θ; a time frequency mask processing section for generating an output sound frequency conversion spectrum ^StC from the time frequency mask ^GtC and the observation signal frequency conversion spectrum XtC; an inverse frequency conversion section for generating distortion output sound ^StC from the output sound frequency conversion spectrum ^StC by using inverse frequency conversion defined by the real number; and a superposing and adding section for generating output sound ^σt from the distortion output sound ^StC.;SELECTED DRAWING: Figure 9;COPYRIGHT: (C)2019,JPO&INPIT
展开▼
机译：解决的问题：提供一种声源增强技术，其使用通过利用由实数定义的频率转换的深层学习估计的时频掩模;解决方案：声源增强处理部分包括：频率转换部分，用于产生观测值利用实数定义的频率转换从观察信号χ t 得到信号频率转换频谱X t C 。时频掩模估计部分，用于根据从观测信号频率提取的声学特征量ψ t 来估计时频掩模^ G t C 神经网络使用参数Θ转换光谱X t C 时间频率掩模处理部分，用于从时间频率掩模^ G t 生成输出声音频率转换频谱^ S t C C 和观测信号频率转换频谱X t C ;逆变频部分，用于从输出声频转换频谱^ S t C <产生失真的输出声^ S t C / Sup>通过使用实数定义的逆频率转换;叠加叠加部分用于从失真输出声音^ S t C 产生输出声音^σ t 。 ;版权：（C）2019，JPO＆INPIT
展开▼
