公开(公告)号：US20240232648A1

公开(公告)日：2024-07-11

申请号：US18537588

申请日：2023-12-12

Applicant: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

Inventor： Hyun-Woo KIM ,  Hwa-Jeon SONG ,  Jeong-Min YANG ,  Byung-Hyun YOO ,  Eui-Sok CHUNG ,  Ran HAN

IPC: G06N3/0985 ,  G06N3/0455 ,  G06N3/088

CPC classification number: G06N3/0985 ,  G06N3/0455 ,  G06N3/088

Abstract: Disclosed herein are a multimodal unsupervised meta-learning method and apparatus. The multimodal unsupervised meta-learning method includes training, by a multimodal unsupervised feature representation learning unit, an encoder configured to extract features of individual single-modal signals from a source multimodal dataset, generating, by a multimodal unsupervised task generation unit, a source task based on the features of individual single-modal signals, deriving, by a multimodal unsupervised learning method derivation unit, a learning method from the source task using the encoder, and training, by a target task performance unit, a model based on the learning method and features extracted from a small number of target datasets by the encoder, thus performing the target task.

公开(公告)号：US20150106097A1

公开(公告)日：2015-04-16

申请号：US14186595

申请日：2014-02-21

Applicant: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

Inventor： Hyun-Woo KIM

IPC: G10L17/00

CPC classification number: G10L25/57 ,  G10L25/78 ,  H04N7/152

Abstract: There is provided a method of determining a main speaker that is performed by a first terminal participating in a distributed telepresence service. The method of determining a main speaker according to an embodiment of the invention includes obtaining first feature information for determining a main speaker from an audio input signal, obtaining second feature information for determining a main speaker of a second terminal from the second terminal participating in the distributed telepresence service, and determining a main speaker terminal for providing a video and an audio of a main speaker who is participating in a telepresence and is speaking based on the first feature information for determining a main speaker and the second feature information for determining a main speaker.

Abstract translation: 提供了一种确定由参与分布式远程呈现服务的第一终端执行的主扬声器的方法。 根据本发明的实施例的确定主扬声器的方法包括从音频输入信号获得用于确定主扬声器的第一特征信息，从参与第二终端的第二终端获得用于确定第二终端的主扬声器的第二特征信息 分布式远程呈现服务，以及基于用于确定主扬声器的第一特征信息和用于确定主扬声器的第二特征信息来确定主扬声器终端，用于提供正在参与远程出席并正在说话的主扬声器的视频和音频 扬声器。

公开(公告)号：US20140324417A1

公开(公告)日：2014-10-30

申请号：US14325475

申请日：2014-07-08

Applicant: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

Inventor： Mi-Suk LEE ,  Heesik YANG ,  Hyun-Woo KIM ,  Jongmo SUNG ,  Hyun-Joo BAE ,  Byung-Sun LEE

IPC: G10L19/02

CPC classification number: G10L19/02 ,  G10L19/0212 ,  G10L19/24

Abstract: Provided are a method and an apparatus for encoding and decoding an audio signal. A method for encoding an audio signal includes receiving a transformed audio signal, dividing the transformed audio signal into a plurality of subbands, performing a first sinusoidal pulse coding operation on the subbands, determining a performance region of a second sinusoidal pulse coding operation among the subbands on the basis of coding information of the first sinusoidal pulse coding operation, and performing the second sinusoidal pulse coding operation on the determined performance region, wherein the first sinusoidal pulse coding operation is performed variably according to the coding information. Accordingly, it is possible to further improve the quality of a synthesized signal by considering the sinusoidal pulse coding of a lower layer when encoding or decoding an audio signal in an upper layer by a layered sinusoidal pulse coding scheme.

Abstract translation: 提供了一种用于对音频信号进行编码和解码的方法和装置。 一种对音频信号进行编码的方法包括接收经变换的音频信号，将变换的音频信号分成多个子带，对子带执行第一正弦脉冲编码操作，确定子带中的第二正弦脉冲编码操作的性能区域 基于第一正弦脉冲编码操作的编码信息，并对所确定的执行区域执行第二正弦脉冲编码操作，其中根据编码信息可变地执行第一正弦脉冲编码操作。 因此，通过考虑在通过分层正弦脉冲编码方案对上层中的音频信号进行编码或解码时下层的正弦脉冲编码，可以进一步提高合成信号的质量。

公开(公告)号：US20130117028A1

公开(公告)日：2013-05-09

申请号：US13661945

申请日：2012-10-26

Applicant: Electronics and Telecommunications Research Institute

Inventor： Hyun-Woo KIM

IPC: G10L19/04

CPC classification number: G10L19/04 ,  G10L19/038 ,  G10L19/12 ,  G10L19/24

Abstract: Disclosed are an apparatus for coding a signal in a communication system including: a coding unit configured to code voice and audio signals based on a code excited linear prediction (CELP) coding method; a residual signal calculation unit configured to calculate residual signals of the voice and audio signals; a frequency transform unit configured to transform the residual signal into a signal in a frequency domain; an energy calculation unit configured to use frequency coefficients of the residual signals to calculate frequency energy of the residual signals; an energy concentration calculation unit configured to calculate energy concentrations of each vector dimension of the residual signals from the frequency energy of the residual signals; and a vector dimension determination unit configured to compare the energy concentrations of each vector dimension to determine targeted vector dimensions of the residual signals.

Abstract translation: 公开了一种用于对通信系统中的信号进行编码的装置，包括：编码单元，被配置为基于代码激励线性预测（CELP）编码方法对语音和音频信号进行编码; 剩余信号计算单元，被配置为计算所述语音和音频信号的残差信号; 频率变换单元，被配置为将所述残差信号变换为频域中的信号; 能量计算单元，被配置为使用所述残余信号的频率系数来计算所述残余信号的频率能量; 能量浓度计算单元，被配置为根据所述残留信号的频率能量计算残留信号的每个向量维数的能量浓度; 以及矢量维度确定单元，被配置为比较每个矢量维度的能量浓度以确定残余信号的目标矢量维度。