Abstract: The technology described herein is directed towards using a wearable passive metasurface (e.g., in the form of a ring) to reflect signals representative of user interaction with computer peripheral devices (e.g., a mouse and keyboard) coupled to a computing device. The reflected signals correspond to user interaction pattern data (user activity profile data) during times of user interaction with the computer. In one implementation, a multi-modal machine learning (ML) classifier processes features of the user interaction pattern data, extracted from time and frequency domain analyses, to recognize habitual interaction patterns of a user, as well as detect any deviations from the user's ordinary interaction behavior, possibly indicating a different user. An interaction pattern that is associated with the user can facilitate ongoing authentication.

Abstract: The technology described herein is directed towards designing and configuring a reconfigurable intelligent surface for deployment, based on a straightforward path-loss model having simplified input variables available to a designer, and having mitigated characterization complexity when compared to other path loss models. The relatively large distance that exists between the feed antenna and a reconfigurable intelligent surface facilitates approximation of certain factors, resulting in a practical solution for design and deployment of a reconfigurable intelligent surface of interest. The input variables include the geometry of the reconfigurable intelligent surface, receiver gain, transmitter gain, and the directivity of the transmitting antenna, which are parameters that are easily available to a designer for deploying a reconfigurable intelligent surface.

Abstract: The disclosure provides a method, apparatus, device, and storage medium for model training optimization. The method includes: determining, for a plurality of operators of a generative machine learning model, resource usage information related to at least one memory optimization operation, where for each memory optimization operation, the resource usage information indicates a resource usage efficiency of performing the memory optimization operation on an operator of the plurality of operators in a training process of the generative machine learning model; determining, based on the resource usage information and available memory resources of a processing device for training the generative machine learning model, a memory optimization strategy for a group of operators of the plurality of operators; and training, by the processing device, the generative machine learning model by applying the memory optimization strategy.

Abstract: The technology described herein is directed towards using a wearable passive metasurface (e.g., configured as a ring) to reflect signals representative of user interaction with computer peripheral devices (e.g., a mouse and keyboard) coupled to a computing device. The reflected signals indicate current user interaction with the device obtained over a timeframe. A short term classifier model set processes features of the user interaction patterns, extracted from time and frequency domain analyses, to recognize activity from the interaction pattern data of a user, as well as detect any deviations from the user's ordinary interaction behavior, possibly indicating a different user. A long term classification model inputs the user interaction pattern data and the recognized activity, to output authentication response data, such as confidence level data, for comparison with a confidence level threshold, to determine via the interaction pattern data whether the user is an authorized user of the computing device.

Abstract: The technology described herein is directed towards designing and deploying a wearable device that includes a passive metasurface of unit cells with a segmented ground plane of electrically separated segments. When the wearable device is worn, e.g., as a ring or wristband that contacts the wearer's skin, the wearer's skin conductivity electrically connects the separated segments to provide a ground plane. With a complete ground plane, the metasurface's unit cells resonate when exposed to a transmitted signal and reflect a distinct signature corresponding to a physical radiation pattern of signals reflected by the activated metasurface. When not worn, the metasurface is deactivated because of the electrically disconnected ground plane segments. Discontinuity parameters of size, position, and quantity of ground discontinuities are iteratively varied during modeling of a metasurface to determine the design with the largest difference in performance characteristics for when the device is worn versus not worn.

Abstract: The technology described herein is directed towards using a wearable passive metasurface (e.g., configured as a ring) to reflect signals representative of user interaction with computer peripheral devices (e.g., a mouse and keyboard) coupled to a computing device. The reflected signals indicate current user interaction with the device obtained over a timeframe. A trained model set processes statistical features of the user interaction patterns, extracted from the current interaction pattern data, to detect any threshold deviation from a prior authorized user's interaction profile data, based on the authorized user's statistically typical interaction behavior on which the model was trained. Any significant deviation indicates a possibly different user.

Abstract: The technology described herein is directed towards metasurfaces arranged with unit cells for narrowband sound absorption, in which the unit cells are based on the principles of Helmholtz resonators. A deeply subwavelength sound absorbing unit-cell is designed and constructed based on a desired resonance frequency. The unit-cell includes a neck portion and air chamber dimensioned to resonate at the desired resonance frequency to inverse phase cancel corresponding narrowband frequencies of incoming sound waves. The unit cells are distributed (e.g., periodically) as part of a metasurface, e.g., positioned proximate to a noise source. As practical examples, the metasurface or multiple metasurfaces can be placed near or wrapped around a computer server or rack of servers to absorb fan noise. The metasurface components (including the unit-cells) can be manufactured using any type of additive manufacturing process such as 3D printing to result in a thin, light-weight, and cost effective noise absorbing metasurface.

Abstract: A method, an apparatus, a system, an electronic device, a computer program product, and a computer-readable storage medium for home care are disclosed. The method includes (i) generating, by a first intelligent agent, a strategy for home care for a user in a current care cycle, based on first historical data and second historical data associated with the home care, the first historical data being generated prior to a current moment in the current care cycle, and the second historical data being generated in a plurality of historical care cycles, (ii) generating, by a second intelligent agent of a plurality of second intelligent agents, a task that may be performed by a tool of home care based on the strategy and the first historical data, and (iii) performing the task through the tool by a third intelligent agent of a plurality of third intelligent agents, which is associated with the second intelligent agent.

Abstract: A target identification method includes: obtaining an image containing a target to be identified, performing feature extraction on the image to obtain image features in the image; and inputting the image features into a target identification network model to obtain an identification result that determines a class to which the target to be identified belongs. The target identification network model includes a loss function that is based on intra-class constraints and inter-class constraints. The intra-class constraints are to constrain an intra-class distance between sample image features of a sample target and a class center of a class to which the sample target belongs, and the inter-class constraints are to constrain inter-class distances between class centers of different classes, and/or inter-class angles between the class centers of different classes.

Abstract: A person re-identification method, a storage medium, and a terminal device are provided. In the method, a preset ratio-based triplet loss function is used as a loss function during training. The ratio-based triplet loss function limits a ratio of a positive sample feature distance to a negative sample feature distance to be less than a preset ratio threshold. The positive sample feature distance is a distance between a reference image feature and a positive sample image feature, and the negative sample feature distance is a distance between the reference image feature and a negative sample image feature. Compared with the existing absolute distance-based triplet loss function, in the case of small inter-class differences and large intra-class differences, the ratio-based triplet loss function can effectively improve the stability of model training, the features extracted by the trained model are more discriminative and robust, thereby improving the accuracy of person re-identification results.

Abstract: A method for pedestrian body part feature extraction is provided, including: performing global feature extraction on a target pedestrian image to obtain a global feature map; learning each of body parts in the global feature map using a self-produced supervision signals-based self-regulated channel attention model to output first channel attention vectors each describing a respective one of the body parts; weighting the first channel attention vectors with the global feature map to obtain a weighted feature map describing the body parts; and extracting body part features of the target pedestrian image from the weighted feature map.

Abstract: Aspects described herein relate to executing address sanitization of instructions for accessing memory. A translation lookaside buffer (TLB) entry indicating a virtual memory page and an address sanitization mode for the virtual memory page can be obtained for a virtual memory address specified in an instruction for accessing memory. It can be determined, based on the address sanitization mode, whether to enable address sanitization for the instruction for accessing memory. A physical memory address corresponding to the virtual memory address indicated in the TLB entry can be accessed based on the instruction for accessing memory.

Abstract: Methods, devices, systems, and techniques for managing conductive structure in semiconductor devices are provided. In one aspect, a semiconductor device includes a memory cell. The memory cell includes a first transistor and a second transistor. The first transistor includes a first semiconductor body extending along a first direction, a first conductive structure that is connected to the first semiconductor body, and a first gate structure that extends along the first direction, where the first semiconductor body is between the first conductive structure and the first gate structure along a second direction perpendicular to the first direction. The second transistor includes a second semiconductor body extending along a first direction, and a second conductive structure, where the second conductive structure is connected to the first conductive structure.

Abstract: The technology described herein is directed towards using a wearable passive metasurface (e.g., configured as a ring) to reflect signals representative of user interaction with computer peripheral devices (e.g., a mouse and keyboard) coupled to a computing device. The reflected signals indicate current user interaction with the device obtained over a timeframe. A short term classifier model set processes features of the user interaction patterns, extracted from time and frequency domain analyses, to recognize activity from the interaction pattern data of a user, as well as detect any deviations from the user's ordinary interaction behavior, possibly indicating a different user. A long term classification model inputs the user interaction pattern data and the recognized activity, to output authentication response data, such as confidence level data, for comparison with a confidence level threshold, to determine via the interaction pattern data whether the user is an authorized user of the computing device.

Abstract: A method for extracting video features may include: obtaining a target video sequence that comprises a number of video frames; performing video frame feature extraction on the target video sequence to obtain video frame features of each of the video frames; performing feature weight calculation on each of the video frame features to obtain the feature weight of each of the video frame features; wherein the feature weight of each of the video frame features is determined by the video frame features of all of the video frames in the target video sequence; and performing feature weighting on each of the video frame features according to the feature weight of each of the video frame features to obtain video features of the target video sequence.

Abstract: The technology described herein is directed towards a passive wearable device with a metasurface, such as in the form of a ring or wristband. The wearable device reflects a transmitted scanning signal to a receiver coupled to a computing device, such as for automatic user proximity detection, user authentication and/or other security purposes. In one implementation, unit cells having resonators in the shape of symmetrical concentric rings around a central disk provide a wide-angle, orientation insensitive ultra-thin and compact wearable device. The symmetry of the unit cells results in the orientation insensitivity, (polarization insensitivity in electromagnetic terms), as well as operating generally independent of the angle of incidence of the transmitted scanning signal. In another implementation, the unit cells for passive wearable devices have resonators in the shape of notched rings, which results in a wide-angle, distance insensitive wearable device that is also significantly orientation insensitive.

Abstract: The technology described herein is directed towards a passive wearable device with a metasurface, such as in the form of a ring or wristband. The wearable device reflects a transmitted scanning signal to a receiver coupled to a computing device, such as for automatic user proximity detection, user authentication and/or other security purposes. In one implementation, unit cells having resonators in the shape of symmetrical concentric rings around a central disk provide a wide-angle, orientation insensitive ultra-thin and compact wearable device. The symmetry of the unit cells results in the orientation insensitivity, (polarization insensitivity in electromagnetic terms), as well as operating generally independent of the angle of incidence of the transmitted scanning signal. In another implementation, the unit cells for passive wearable devices have resonators in the shape of notched rings, which results in a wide-angle, distance insensitive wearable device that is also significantly orientation insensitive.

Abstract: The technology described herein is directed towards using a wearable passive metasurface (e.g., in the form of a ring) to reflect signals representative of user interaction with computer peripheral devices (e.g., a mouse and keyboard) coupled to a computing device. The reflected signals correspond to user interaction pattern data (user activity profile data) during times of user interaction with the computer. In one implementation, a multi-modal machine learning (ML) classifier processes features of the user interaction pattern data, extracted from time and frequency domain analyses, to recognize habitual interaction patterns of a user, as well as detect any deviations from the user's ordinary interaction behavior, possibly indicating a different user. An interaction pattern that is associated with the user can facilitate ongoing authentication.

Abstract: A method for obtaining a feature extraction model, a method for human fall detection and a terminal device are provided. The method for human fall detection includes: inputting a human body image into a feature extraction model for feature extraction to obtain a target image feature; in response to a distance between the target image feature and a pre-stored mean value of standing category image features being greater than or equal to a preset distance threshold, determining that the human body image is a human falling image; and in response to the distance being less than the preset distance threshold, determining that the human body image is a human standing image. The feature extraction model is obtained based on constraint training to aggregate standing category image features and separate falling category image features from the standing category image features.

Abstract: A method to predict and fit an APC of a LEO satellite considering attitude information is provided, which includes: transforming a spacecraft body-fixed coordinate system, obtaining a transformation matrix from orbital coordinate system to transformed spacecraft body-fixed coordinate system, obtaining rotation angles during a rotation process, obtaining rotation angles in orbit determination time period to predict rotation angles in orbit prediction time period, obtaining a transformation matrix from spacecraft body-fixed coordinate system to Earth-fixed coordinate system of each prediction moment, obtaining a correction vector from a CoM to a phase center of a downlink signal antenna in the Earth-fixed coordinate system at each prediction moment, and obtaining an orbit of the phase center of the downlink signal antenna of a LEO satellite, and performing an ephemeris parameter fitting on the orbit of the phase center of the downlink signal antenna of the LEO satellite in a fitting time period.