Abstract: A method for training a speech conversion model personalized for a target speaker with atypical speech includes obtaining a plurality of transcriptions in a set of spoken training utterances and obtaining a plurality of unspoken training text utterances. Each spoken training utterance is spoken by a target speaker associated with atypical speech and includes a corresponding transcription paired with a corresponding non-synthetic speech representation. The method also includes adapting, using the set of spoken training utterances, a text-to-speech (TTS) model to synthesize speech in a voice of the target speaker and that captures the atypical speech. For each unspoken training text utterance, the method also includes generating, as output from the adapted TTS model, a synthetic speech representation that includes the voice of the target speaker and that captures the atypical speech. The method also includes training the speech conversion model based on the synthetic speech representations.

Abstract: An image processing method for an augmented reality device, an augmented reality device and a storage medium are provided. The image processing method for the augmented reality device includes: tracking lines of sight of a left eye and a right eye to determine a position of a binocular watching range in a three-dimensional space; shooting an image of an external environment where the augmented reality (AR) device is located, and acquiring environment map point cloud data based on a Simultaneous Localization and Mapping (SLAM) algorithm; and updating the environment map point cloud data within the binocular watching range in response to detecting an update of the image of the external environment.

Abstract: Disclosed are a light-emitting device, a template of the light-emitting device and preparation methods thereof. The template of the light-emitting device comprises a substrate; a GaN-based semiconductor layer and a mask layer provided on the substrate, where the mask layer comprises a plurality of mask openings provided at intervals, and the plurality of mask openings are filled with the GaN-based semiconductor layer; and a sacrificial layer provided on a surface of the GaN-based semiconductor layer away from the substrate and located in the plurality of mask openings provided at intervals.

Abstract: The application discloses a method and a device for calculating parameters of a gaze-point calculation model. The method includes: displaying multiple calibration points in sequence on a device screen, and collecting multiple pupil-coordinates data corresponding to each of at least one calibration point of the multiple calibration points; for each calibration point of the at least one calibration point, screening the multiple pupil-coordinates data corresponding to the calibration point, to obtain at least one pupil-coordinates calibration data corresponding to the calibration point; calculating multiple parameters of the gaze-point calculation model according to coordinates data of each calibration point and the at least one pupil-coordinates calibration data corresponding to the calibration point.

Abstract: Disclosed is a method for preparing a substrate relate to the field of semiconductors. The method comprises the following steps: S1, providing a reaction container in which a base substrate is mounted; S2, conducting a metal source into the reaction container, and forming a thin film layer on a surface of the base substrate, wherein a part of a surface of the base substrate is covered by the thin film layer, so that the base substrate is provided with an exposed surface that is not covered by the thin film layer; and S3, conducting a corrosive gas into the reaction container to form one or more recessed holes in at least a part of the exposed surface.

Abstract: An embodiment of an electronic apparatus may comprise one or more substrates and a controller coupled to the one or more substrates, the controller including circuitry to identify failed memory regions in a memory by a rank, bank, and device associated with the failed memory region, and provide recovery for failed memory regions in three or more banks of a first rank of the memory or three or more devices of the first rank of the memory by virtual lock step device data correction with one or more other ranks of the memory. Other embodiments are disclosed and claimed.

Abstract: This application provides a method and an apparatus for determining a port orientation. The method includes: A first network device receives, through a first port, a protocol data unit PDU sent by a second network device. The first network device determines, based on the PDU, that the first port is an upstream port or a downstream port. In this embodiment of this application, the first network device may determine, based on the PDU sent by the second network device, whether the first port is the upstream port or the downstream port, to provide a basis for subsequently helping an SDN controller reduce load.

Abstract: A method for model deployment, a terminal device, and a non-transitory computer-readable storage medium are provided. The method includes the following. A to-be-deployed model and an input/output description file of the to-be-deployed model are obtained. Output verification is performed on the to-be-deployed model based on the input/output description file. If the output verification of the to-be-deployed model passes, an inference service resource is determined from multiple running environments and the inference service resource is allocated to the to-be-deployed model. An inference parameter value of executing an inference service by the to-be-deployed model based on the inference service resource is determined. A resource configuration file and an inference service interface of the to-be-deployed model are generated according to the inference service resource, if the inference parameter value is greater than or equal to a preset inference parameter threshold.

Abstract: A method for training a speech conversion model personalized for a target speaker with atypical speech includes obtaining a plurality of transcriptions in a set of spoken training utterances and obtaining a plurality of unspoken training text utterances. Each spoken training utterance is spoken by a target speaker associated with atypical speech and includes a corresponding transcription paired with a corresponding non-synthetic speech representation. The method also includes adapting, using the set of spoken training utterances, a text-to-speech (TTS) model to synthesize speech in a voice of the target speaker and that captures the atypical speech. For each unspoken training text utterance, the method also includes generating, as output from the adapted TTS model, a synthetic speech representation that includes the voice of the target speaker and that captures the atypical speech. The method also includes training the speech conversion model based on the synthetic speech representations.

Abstract: According to some embodiments, a multi-layer speech recognition transcript post processing system may include a data-driven, statistical layer associated with a trained automatic speech recognition model that selects an initial transcript. A rule-based layer may receive the initial transcript from the data-driven, statistical layer and execute at least one pre-determined rule to generate a first modified transcript. A machine learning approach layer may receive the first modified transcript from the rule-based layer and perform a neural model inference to create a second modified transcript. A human editor layer may receive the second modified transcript from the machine learning approach layer along with an adjustment from at least one human editor. The adjustment may create, in some embodiments, a final transcript that may be used to fine-tune the data-driven, statistical layer.

Abstract: The present disclosure discloses an engineering machinery equipment, and a method, system, and storage medium for operation trajectory planning thereof, and relates to the field of artificial intelligence, automatic control, and engineering machinery technologies.

Abstract: The present disclosure discloses an engineering machinery equipment, and a method, system, and storage medium for safety control thereof, and relates to the field of artificial intelligence, automatic control, and engineering machinery technologies. A method can include: acquiring spatial sensing data of a work area; performing obstacle detection based on the spatial sensing data to determine a position of an obstacle within the work area; determining a safe working range of a mechanical structural component of the engineering machinery equipment based on the position of the obstacle within the work area; and controlling a working range of the mechanical structural component of the engineering machinery equipment based on the safe working range.

Abstract: This application provides a virtual channel setting method and the method includes: a' receiving device receives a first data packet of a first data flow from a sending device, where the first data packet includes an identifier of a first virtual channel corresponding to the first data flow; the receiving device selects a second virtual channel for the first data flow from a first virtual channel group corresponding to the first data flow, where the second virtual channel is different from the first virtual channel; and the receiving device sends an acknowledgement packet to the sending device, where the acknowledgement packet includes an identifier of the second virtual channel, and the second virtual channel is used by the sending device to update a virtual channel corresponding to the first data flow.

Abstract: A method of manufacturing a lead assembly of a cryogenic system is provided. The method includes developing a three-dimensional (3D) model of a heat exchanger. The heat exchanger includes a plurality of channels extending longitudinally through the heat exchanger from the first end to the second end, the plurality of channels forming a plurality of thermal surfaces within the heat exchanger, the heat exchanger having a transverse cross section. The method further includes modifying the 3D model by at least one of reducing an area of the cross section and increasing the plurality of thermal surfaces. The method also includes additively manufacturing the heat exchanger using an electrically-conductive and thermally-conductive material according to the modified 3D model. Further, the method includes providing a high temperature superconductor (HTS) assembly that includes an HTS strip, and connecting the HTS assembly to the heat exchanger at the second end of the heat exchanger.

Abstract: The invention includes a method including the steps of obtaining a plurality of images, each of the images in the plurality having at least one corresponding region, generating a merged image, the merged image also having the corresponding region. The step of generating includes selecting an image source from the plurality of images to source image data for the corresponding region in the merged image by comparing attributes of the corresponding regions of the plurality of images to identify the image source having preferred attributes.

Abstract: Embodiments of the present disclosure disclose a method, apparatus, device, and storage medium for controlling a guide robot, and relate to the field of artificial intelligence, robots, and multi-sensor fusion technologies. A specific embodiment of the method includes: acquiring a state of the guide robot, a state of a user, and a position of an obstacle; generating a state update equation for a combined system of the guide robot and the user based on the state of the guide robot and the state of the user; generating a collision-free global path based on the position of the obstacle; generating a control command based on the state update equation for the combined system and the collision-free global path; and driving the guide robot to move based on the control command.

Abstract: A surface roughening method includes the following steps: preparing a first epitaxial layer of a three-dimensional island shape growth over a light emitting structure; and preparing a discontinuous second epitaxial layer over the first epitaxial layer. The surface roughening method provided in the present application is simple and convenient, and improves the efficiency. In addition to the epitaxial growth process, it is not necessary to use an additional process such as wet etching, photonic crystal and other processes to further process the surface of the epitaxial layer, and the method may be implemented by means of one process in a same reaction equipment.

Abstract: A transcription of audio speech included in electronic content associated with a meeting is created by an ASR model trained on speech-to-text data. The transcription is post-processed by modifying text included in the transcription, for example, by modifying punctuation, grammar, or formatting introduced by the ASR model and by changing or omitting one or more words that were included in both the audio speech and the transcription. After the transcription is post-processed, output based on the post-processed transcription is generated in the form of a meeting summary and/or template.

Abstract: Methods are provided for identifying donors of fecal matter that can improve a subject's response to a checkpoint inhibitor in the treatment of cancer. Methods and compositions are also provided for using donated fecal matter in the treatment of cancer.

Abstract: The present application provides an intelligent cloud traffic control system, which has a traffic data recalling and transmission architecture arranged hierarchically from bottom to top, and which includes: terminal devices, first-level cloud devices, second-level cloud devices, . . . ith-level cloud devices; wherein the ith-level cloud devices include a first cloud device, . . . a jth cloud device; the ith-level cloud devices further include: a first cache cloud, . . . a kth cache cloud arranged corresponding to the jth cloud device; wherein the jth cloud device stores a jth cloud device traffic data set, and the kth cache cloud stores a kth characteristic traffic data set; and data in the kth characteristic traffic data set has a higher priority of being recalled than data in the jth cloud device traffic data set.