公开(公告)号：US20050278172A1

公开(公告)日：2005-12-15

申请号：US10869467

申请日：2004-06-15

申请人： Kazuhito Koishida ,  Feng Zhuge ,  Hosam Khalil ,  Tian Wang ,  Wei-ge Chen

发明人： Kazuhito Koishida ,  Feng Zhuge ,  Hosam Khalil ,  Tian Wang ,  Wei-ge Chen

IPC分类号： G10L15/20 ,  G10L21/02

CPC分类号： G10L21/0208 ,  G10L21/0232

摘要： A gain-constrained noise suppression for speech more precisely estimates noise, including during speech, to reduce musical noise artifacts introduced from noise suppression. The noise suppression operates by applying a spectral gain G(m, k) to each short-time spectrum value S(m, k) of a speech signal, where m is the frame number and k is the spectrum index. The spectrum values are grouped into frequency bins, and a noise characteristic estimated for each bin classified as a “noise bin.” An energy parameter is smoothed in both the time domain and the frequency domain to improve noise estimation per bin. The gain factors G(m, k) are calculated based on the current signal spectrum and the noise estimation, then smoothed before being applied to the signal spectral values S(m, k). First, a noisy factor is computed based on a ratio of the number of noise bins to the total number of bins for the current frame, where a zero-valued noisy factor means only using constant gain for all the spectrum values and noisy factor of one means no smoothing at all. Then, this noisy factor is used to alter the gain factors, such as by cutting off the high frequency components of the gain factors in the frequency domain.

摘要翻译： 用于语音的增益约束噪声抑制更精确地估计包括在语音期间的噪声，以减少从噪声抑制引入的音乐噪声伪像。 通过对语音信号的每个短时间频谱值S（m，k）应用频谱增益G（m，k）来进行噪声抑制，其中m是帧号，k是频谱索引。 频谱值被分组成频率仓，并且对于被分类为“噪声仓”的每个仓估计的噪声特性。 能量参数在时域和频域均被平滑，以改善每个bin的噪声估计。 基于当前信号频谱和噪声估计来计算增益因子G（m，k），然后在施加到信号频谱值S（m，k）之前进行平滑处理。 首先，基于噪声箱数与当前帧的总数的比率来计算噪声因子，其中零值噪声因子意味着仅对所有频谱值使用恒定增益并且噪声因子为1 意味着没有平滑。 然后，这种噪声因子用于改变增益因子，例如通过切断频域中增益因子的高频分量。

公开(公告)号：US07454332B2

公开(公告)日：2008-11-18

申请号：US10869467

申请日：2004-06-15

申请人： Kazuhito Koishida ,  Feng Zhuge ,  Hosam A. Khalil ,  Tian Wang ,  Wei-ge Chen

发明人： Kazuhito Koishida ,  Feng Zhuge ,  Hosam A. Khalil ,  Tian Wang ,  Wei-ge Chen

IPC分类号： G10L21/02 ,  G10L19/14

CPC分类号： G10L21/0208 ,  G10L21/0232

摘要： A gain-constrained noise suppression for speech more precisely estimates noise, including during speech, to reduce musical noise artifacts introduced from noise suppression. The noise suppression operates by applying a spectral gain G(m, k) to each short-time spectrum value S(m, k) of a speech signal, where m is the frame number and k is the spectrum index. The spectrum values are grouped into frequency bins, and a noise characteristic estimated for each bin classified as a “noise bin.” An energy parameter is smoothed in both the time domain and the frequency domain to improve noise estimation per bin. The gain factors G(m, k) are calculated based on the current signal spectrum and the noise estimation, then smoothed before being applied to the signal spectral values S(m, k). First, a noisy factor is computed based on a ratio of the number of noise bins to the total number of bins for the current frame, where a zero-valued noisy factor means only using constant gain for all the spectrum values and noisy factor of one means no smoothing at all. Then, this noisy factor is used to alter the gain factors, such as by cutting off the high frequency components of the gain factors in the frequency domain.

摘要翻译： 用于语音的增益约束噪声抑制更精确地估计包括在语音期间的噪声，以减少从噪声抑制引入的音乐噪声伪像。 通过对语音信号的每个短时间频谱值S（m，k）应用频谱增益G（m，k）来进行噪声抑制，其中m是帧号，k是频谱索引。 频谱值被分组成频率仓，并且对于被分类为“噪声仓”的每个仓估计的噪声特性。 能量参数在时域和频域均被平滑，以改善每个bin的噪声估计。 基于当前信号频谱和噪声估计来计算增益因子G（m，k），然后在施加到信号频谱值S（m，k）之前进行平滑处理。 首先，基于噪声箱数与当前帧的总数的比率来计算噪声因子，其中零值噪声因子意味着仅对所有频谱值使用恒定增益并且噪声因子为1 意味着没有平滑。 然后，这种噪声因子用于改变增益因子，例如通过切断频域中增益因子的高频分量。

公开(公告)号：US07840504B2

公开(公告)日：2010-11-23

申请号：US11805424

申请日：2007-05-22

申请人： Shaohua Sun ,  Feng Zhuge ,  Sandy A. Napel

发明人： Shaohua Sun ,  Feng Zhuge ,  Sandy A. Napel

IPC分类号： G06F15/18

CPC分类号： G06N5/02

摘要： A Learning Enhanced Simulated Annealing (LESA) method is provided. Based on a Simulated Annealing (SA) framework, this method adds a Knowledge Base (KB) initialized at the beginning of the search and updated at each iteration, which memorizes a portion of the search history and guides the further search through a KB trial generator. The basic idea of LESA is that its search history is stored in a KB, and a KB trial generator extracts information from it and uses it to generate a new trial. The next move of the search is the weighted sum of the trial generated by the KB trial generator and the trial generated by the usual SA trial generator. The knowledge base is then updated after each search iteration.

摘要翻译： 提供学习增强模拟退火（LESA）方法。 基于模拟退火（SA）框架，该方法添加了在搜索开始时初始化的知识库（KB），并在每次迭代中更新，这些记录了搜索历史的一部分，并通过KB试用生成器指导进一步的搜索 。 LESA的基本思想是其搜索历史记录存储在KB中，KB试验生成器从中提取信息并使用它来生成新的试用版。 搜索的下一步是由KB试验发生器产生的试验和由通常的SA试验发生器产生的试验的加权和。 然后在每次搜索迭代后更新知识库。

公开(公告)号：US20070299801A1

公开(公告)日：2007-12-27

申请号：US11805424

申请日：2007-05-22

申请人： Shaohua Sun ,  Feng Zhuge ,  Sandy Napel

发明人： Shaohua Sun ,  Feng Zhuge ,  Sandy Napel

IPC分类号： G06F17/00

CPC分类号： G06N5/02

摘要： A Learning Enhanced Simulated Annealing (LESA) method is provided. Based on a Simulated Annealing (SA) framework, this method adds a Knowledge Base (KB) initialized at the beginning of the search and updated at each iteration, which memorizes a portion of the search history and guides the further search through a KB trial generator. The basic idea of LESA is that its search history is stored in a KB, and a KB trial generator extracts information from it and uses it to generate a new trial. The next move of the search is the weighted sum of the trial generated by the KB trial generator and the trial generated by the usual SA trial generator. The knowledge base is then updated after each search iteration.

摘要翻译： 提供学习增强模拟退火（LESA）方法。 基于模拟退火（SA）框架，该方法添加了在搜索开始时初始化的知识库（KB），并在每次迭代中更新，这些记录了搜索历史的一部分，并通过KB试用版引导进一步的搜索 。 LESA的基本思想是其搜索历史记录存储在KB中，KB试验生成器从中提取信息并使用它来生成新的试用版。 搜索的下一步是由KB试验发生器产生的试验和由通常的SA试验发生器产生的试验的加权和。 然后在每次搜索迭代后更新知识库。