Inventor
YIFAN GONG
Inventor
US
Overview
Patents· 30 as inventor
Public. Assignee & technology derived from the patent record. Each patent links to its LexDana page.
G10Lprimary
G10LSpeech analysis techniques or speech synthesis46%
G06NComputing arrangements based on specific computational models28%
G06FElectric digital data processing16%
G06KGraphical data reading3%
H04RLoudspeakers, microphones, gramophone pick-ups or like acoustic electromechanical transducers3%
G09BEducational or demonstration appliances2%
H04LTransmission of digital information, e.g. telegraphic communication2%
PatentTitleYear
12,205,596Generating and using text-to-speech data for speech recognition models202512,141,080Communication method, related computing system and storage medium for selecting a communication channel between modules202412,086,704Machine learning model with depth processing units202412,014,728Dynamic combination of acoustic model states202411,915,686Speaker adaptation for attention-based encoder-decoder202411,862,144Augmented training data for end-to-end models202411,823,702Condition-invariant feature extraction network202311,798,535On-device custom wake word detection202311,790,891Wake word selection assistance architectures and methods202311,776,548Convolutional neural network with phonetic attention for speaker verification202311,735,190Attentive adversarial domain-invariant training202311,705,117Adaptive batching to reduce recognition latency202311,676,006Universal acoustic modeling using neural mixture models202311,657,799Pre-training with alignments for recurrent neural network transducer based end-to-end speech recognition202311,631,399Layer trajectory long short-term memory with future context202311,620,992Automated speech recognition confidence classifier202311,587,569Generating and using text-to-speech data for speech recognition models202311,586,930Conditional teacher-student learning for model training202311,563,784Caption assisted calling to maintain connection in challenging network conditions202311,562,745Sequence-to-sequence speech recognition with latency threshold202311,527,238Internal language model for E2E models202211,429,860Learning student DNN via output distribution202211,276,410Convolutional neural network with phonetic attention for speaker verification202211,244,673Streaming contextual unidirectional models202211,232,782Speaker adaptation for attention-based encoder-decoder202211,222,622Wake word selection assistance architectures and methods202211,217,265Condition-invariant feature extraction network202211,210,565Machine learning model with depth processing units202111,183,178Adaptive batching to reduce recognition latency202111,158,305Online verification of custom wake word2021Career & activity timeline
Assembled from patent assignee/location history and declaration credentials.
2021–2025Patents assigned to MICROSOFT TECHNOLOGY LICENSING LLC (29)· patent assignee history
2021–2025Based in US· patent location history
2024Patents assigned to BEIJING TUSEN ZHITU TECH CO LTD (1)· patent assignee history
2024Based in CN· patent location history
Inventor activity
Assignee and location history derived from the patent record.
Assignees
| Assignee | Patents | Span |
|---|---|---|
| MICROSOFT TECHNOLOGY LICENSING LLC | 29 | 2021–2025 |
| BEIJING TUSEN ZHITU TECH CO LTD | 1 | 2024 |
Locations
| Location | Patents | Span |
|---|---|---|
| US | 29 | 2021–2025 |
| CN | 1 | 2024 |