-
公开(公告)号：US20230326452A1
公开(公告)日：2023-10-12
申请号：US18189145
申请日：2023-03-23
Applicant: Sony Interactive Entertainment Inc.
Inventor： Zhenhao Ge , Lakshmish Kaushik , Saket Kumar , Masanori Omote
CPC classification number: G10L15/063 , G10L13/02
Abstract: Techniques are described for training neural networks on variable length datasets. The numeric representation of the length of each training sample is randomly perturbed to yield a pseudo-length, and the samples sorted by pseudo-length to achieve lower zero padding rate (ZPR) than completely randomized batching (thus saving computation time) yet higher randomness than strictly sorted batching (thus achieving better model performance than strictly sorted batching).
-
公开(公告)号：US20210129031A1
公开(公告)日：2021-05-06
申请号：US16674827
申请日：2019-11-05
Applicant: Sony Interactive Entertainment Inc.
Inventor： Joshua M. Eads , Matthew D. Bennett , Brendan Matera Rehon , Mahdi Azmandian , Zhenhao Ge , Todd Tokubo
Abstract: The performance of a player of a computer game is noted and the player accorded a latency handicap based thereon. The latency handicap is used to slow down play of the computer game, preferably only during times of high player activity. The latency handicap can be reduced over time or owing to improvement in the player's performance.
-
公开(公告)号：US11615782B2
公开(公告)日：2023-03-28
申请号：US17107362
申请日：2020-11-30
Applicant: Sony Interactive Entertainment Inc.
Inventor： Zhenhao Ge , Lakshmish Kaushik , Saket Kumar , Masanori Omote
Abstract: Techniques are described for training neural networks on variable length datasets. The numeric representation of the length of each training sample is randomly perturbed to yield a pseudo-length, and the samples sorted by pseudo-length to achieve lower zero padding rate (ZPR) than completely randomized batching (thus saving computation time) yet higher randomness than strictly sorted batching (thus achieving better model performance than strictly sorted batching).
-
公开(公告)号：US11915685B2
公开(公告)日：2024-02-27
申请号：US18189145
申请日：2023-03-23
Applicant: Sony Interactive Entertainment Inc.
Inventor： Zhenhao Ge , Lakshmish Kaushik , Saket Kumar , Masanori Omote
CPC classification number: G10L15/063 , G10L13/02
Abstract: Techniques are described for training neural networks on variable length datasets. The numeric representation of the length of each training sample is randomly perturbed to yield a pseudo-length, and the samples sorted by pseudo-length to achieve lower zero padding rate (ZPR) than completely randomized batching (thus saving computation time) yet higher randomness than strictly sorted batching (thus achieving better model performance than strictly sorted batching).
-
公开(公告)号：US20220148569A1
公开(公告)日：2022-05-12
申请号：US17107362
申请日：2020-11-30
Applicant: Sony Interactive Entertainment Inc.
Inventor： Zhenhao Ge , Lakshmish Kaushik , Saket Kumar , Masanori Omote
Abstract: Techniques are described for training neural networks on variable length datasets. The numeric representation of the length of each training sample is randomly perturbed to yield a pseudo-length, and the samples sorted by pseudo-length to achieve lower zero padding rate (ZPR) than completely randomized batching (thus saving computation time) yet higher randomness than strictly sorted batching (thus achieving better model performance than strictly sorted batching).
-
Patent Agency Ranking
公开(公告)号：US11217245B2
公开(公告)日：2022-01-04
申请号：US16555616
申请日：2019-08-29
Applicant: Sony Interactive Entertainment Inc.
Inventor： Lakshmish Kaushik , Zhenhao Ge
Abstract: A wake-up word for a digital assistant may be specified by a user to trigger the digital assistant to respond to the wake-up word, with the user providing one or more initial pronunciations of the wake-up word. The wake-up word may be unique, or at least not determined beforehand by a device manufacturer or developer of the digital assistant. The initial pronunciation(s) of the keyword may then be augmented with other potential pronunciations of the wake-up word that might be provided in the future, and those other potential pronunciations may then be pruned down to a threshold number of other potential pronunciations. One or more recordings of the initial pronunciation(s) of the wake-up may then be used to train a phoneme recognizer model to better recognize future instances of the wake-up word being spoken by the user or another person using the initial pronunciation or other potential pronunciations.
-
公开(公告)号：US12290753B2
公开(公告)日：2025-05-06
申请号：US16674827
申请日：2019-11-05
Applicant: Sony Interactive Entertainment Inc.
Inventor： Joshua M. Eads , Matthew D. Bennett , Brendan Matera Rehon , Mahdi Azmandian , Zhenhao Ge , Todd Tokubo
Abstract: The performance of a player of a computer game is noted and the player accorded a latency handicap based thereon. The latency handicap is used to slow down play of the computer game, preferably only during times of high player activity. The latency handicap can be reduced over time or owing to improvement in the player's performance.
-
公开(公告)号：US11790912B2
公开(公告)日：2023-10-17
申请号：US17567873
申请日：2022-01-03
Applicant: Sony Interactive Entertainment Inc.
Inventor： Lakshmish Kaushik , Zhenhao Ge , Xiaoyu Liu
CPC classification number: G10L15/22 , G06F40/242 , G10L15/02 , G10L15/063 , G10L15/16 , G10L2015/025 , G10L2015/088
Abstract: A wake-up word for a digital assistant may be specified by a user to trigger the digital assistant to respond to the wake-up word, with the user providing one or more initial pronunciations of the wake-up word. The wake-up word may be unique, or at least not determined beforehand by a device manufacturer or developer of the digital assistant. The initial pronunciation(s) of the keyword may then be augmented with other potential pronunciations of the wake-up word that might be provided in the future, and those other potential pronunciations may then be pruned down to a threshold number of other potential pronunciations. One or more recordings of the initial pronunciation(s) of the wake-up may then be used to train a phoneme recognizer model to better recognize future instances of the wake-up word being spoken by the user or another person using the initial pronunciation or other potential pronunciations.
-
公开(公告)号：US20220130384A1
公开(公告)日：2022-04-28
申请号：US17567873
申请日：2022-01-03
Applicant: Sony Interactive Entertainment Inc.
Inventor： Lakshmish Kaushik , Zhenhao Ge
IPC: G10L15/22 , G10L15/02 , G10L15/16 , G10L15/06 , G06F40/242
Abstract: A wake-up word for a digital assistant may be specified by a user to trigger the digital assistant to respond to the wake-up word, with the user providing one or more initial pronunciations of the wake-up word. The wake-up word may be unique, or at least not determined beforehand by a device manufacturer or developer of the digital assistant. The initial pronunciation(s) of the keyword may then be augmented with other potential pronunciations of the wake-up word that might be provided in the future, and those other potential pronunciations may then be pruned down to a threshold number of other potential pronunciations. One or more recordings of the initial pronunciation(s) of the wake-up may then be used to train a phoneme recognizer model to better recognize future instances of the wake-up word being spoken by the user or another person using the initial pronunciation or other potential pronunciations.
-
公开(公告)号：US20210065699A1
公开(公告)日：2021-03-04
申请号：US16555616
申请日：2019-08-29
Applicant: Sony Interactive Entertainment Inc.
Inventor： Lakshmish Kaushik , Zhenhao Ge
Abstract: A wake-up word for a digital assistant may be specified by a user to trigger the digital assistant to respond to the wake-up word, with the user providing one or more initial pronunciations of the wake-up word. The wake-up word may be unique, or at least not determined beforehand by a device manufacturer or developer of the digital assistant. The initial pronunciation(s) of the keyword may then be augmented with other potential pronunciations of the wake-up word that might be provided in the future, and those other potential pronunciations may then be pruned down to a threshold number of other potential pronunciations. One or more recordings of the initial pronunciation(s) of the wake-up may then be used to train a phoneme recognizer model to better recognize future instances of the wake-up word being spoken by the user or another person using the initial pronunciation or other potential pronunciations.
-
-
-
-
-
-
-
-
-
Search Results
10
records
(took 0.023 seconds)
