公开(公告)号：WO2016099301A1

公开(公告)日：2016-06-23

申请号：PCT/PL2014/000142

申请日：2014-12-17

Applicant: INTEL CORPORATION ,  MALINOWSKI, Lukasz, M.

Inventor： MALINOWSKI, Lukasz, M. ,  MAJCHER, Piotr ,  STEMMER, Georg ,  ROZEN, Piotr ,  HOFER, Joachim ,  BAUER, Josef, G.

IPC: G10L15/34 ,  G10L15/08

CPC classification number: G10L15/193 ,  G10L15/083 ,  G10L15/22 ,  G10L15/34

Abstract: A system, article, and method of automatic speech recognition using parallel processing weighted finite state transducer-based speech decoding.

Abstract translation: 一种使用并行处理加权有限状态传感器的语音解码的自动语音识别的系统，文章和方法。

公开(公告)号：WO2016048486A1

公开(公告)日：2016-03-31

申请号：PCT/US2015/045750

申请日：2015-08-18

Applicant: INTEL CORPORATION

Inventor： BAUER, Josef ,  ROZEN, Piotr ,  STEMMER, Georg

IPC: G10L15/16

CPC classification number: G10L15/16 ,  G10L15/02 ,  G10L15/08 ,  G10L15/12

Abstract: Techniques related to implementing neural networks for speech recognition systems are discussed. Such techniques may include implementing frame skipping with approximated skip frames and/or distances on demand such that only those outputs needed by a speech decoder are provided via the neural network or approximation techniques.

Abstract translation: 讨论了与语音识别系统实现神经网络有关的技术。 这样的技术可以包括实现具有近似跳过帧和/或按需的距离的跳帧，使得仅通过神经网络或近似技术提供语音解码器所需的那些输出。

公开(公告)号：WO2016039651A1

公开(公告)日：2016-03-17

申请号：PCT/PL2014/050053

申请日：2014-09-09

Applicant: INTEL CORPORATION

Inventor： ROZEN, Piotr ,  STEMMER, Georg

IPC: G06N3/063 ,  G10L15/16

CPC classification number: G06N3/08 ,  G06F7/483 ,  G06N3/04 ,  G06N3/063 ,  G10L15/02 ,  G10L15/16

Abstract: Techniques related to implementing neural networks for speech recognition systems are discussed. Such techniques may include processing a node of the neural network by determining a score for the node as a product of weights and inputs such that the weights are fixed point integer values, applying a correction to the score based a correction value associated with at least one of the weights, and generating an output from the node based on the corrected score.

Abstract translation: 讨论了与语音识别系统实现神经网络相关的技术。 这样的技术可以包括通过将节点的得分确定为权重和输入的乘积来处理神经网络的节点，使得权重是固定点整数值，对基于与至少一个相关联的校正值进行校正 的权重，并且基于校正得分从节点生成输出。