31 results about "Multi stream" patented technology

A device includes one or more processors configured to detect single-stream data and generate multi-stream augmented data that includes one or more modified versions of the single-stream data. The one or more processors are configured to process the multi-stream augmented data to generate multiple output channels. The one or more processors are also configured to reduce the multiple output channels to produce single-stream output data.

Disclosed embodiments relate to Multi-Player Playback Devices. In some embodiments, the Multi-Player Playback Device is configured to selectively operate in one of a plurality of operating modes. In each operating mode, the Multi-Player Playback Device is configured to implement a multi-stage audio processing procedure comprising: (i) routing incoming audio streams to mixer inputs of a mixer based on a current operating mode in which the Multi-Player Playback Device is operating; (ii) for one or more audio streams routed to an individual mixer input, (a) generating a mixed stream based on the one or more audio streams routed to the mixer input, and (b) routing the mixed stream to at least one audio output of the Multi-Player Playback Device based at least in part on a channel map associated with the current operating mode in which the Multi-Player Playback Device is operating.

A multi-stream GPU video memory multiplexing method oriented to tensor structure equivalence is characterized in that a video memory block of a computational stream is placed in a video memory shared pool after being released, and the method comprises the following steps: extracting structure information of a tensor, including dimension, data type and step length layout, and generating a unique structure identifier based on the structure information of the tensor; a mapping table is constructed and maintained, a plurality of key-value pairs are stored in the mapping table, keys are unique structure identifiers, values are video memory addresses, and video memory blocks are located in a video memory shared pool; when the tensor video memory is allocated, whether a video memory address which is matched with the unique structure identifier and is not occupied exists in a mapping table or not is inquired in combination with the current execution context identifier, and if yes, the pointer is adjusted to point to the video memory address, and the unoccupied video memory is allocated for the tensor; and if not, applying for a video memory address and updating the mapping table. According to the method, on the premise that an original network structure and execution logic are not changed, a large number of tensor spaces with non-overlapped life cycles can be reused, and video memory peak value occupation is remarkably reduced.

A multi-stream GPU memory multiplexing method for tensor structure equivalence, after the memory block of a computing stream is released, it is put into a memory sharing pool, the method comprises the following steps: extracting the structure information of a tensor, including dimension, data type and step layout, generating a unique structure identifier based on the structure information of the tensor; constructing and maintaining a mapping table, the mapping table stores a plurality of key-value pairs, the key is a unique structure identifier, and the value is a memory address, the memory block is located in the memory sharing pool; when allocating the memory of the tensor, in combination with the current execution context identifier, query whether there is a memory address in the mapping table that matches the unique structure identifier and is not occupied, if so, adjust the pointer to point to the memory address, and allocate the unoccupied memory to the tensor; if not, apply for a memory address and update the mapping table. The method can reuse a large number of non-overlapping tensor spaces without changing the original network structure and execution logic, and significantly reduces the peak memory occupancy.

In some embodiments, a system can include a first stream to be opened with an arbiter device, the first stream associated with a first set of capabilities, a second stream to be opened with the arbiter device, the second stream associated with a second set of capabilities, and a wireless network architecture that determines a network messaging schedule based on the first set of capabilities and the second set of capabilities. The network messaging schedule can be sent to one or more client devices associated with the first stream or the second stream.

This invention belongs to the field of computer vision and intelligent environmental perception technology, specifically relating to a method for anomaly detection and intensity estimation of sandstorms based on event cameras. It includes constructing and training an estimation model based on a multi-stream regression framework, and using the trained estimation model to predict sandstorm concentration. The estimation model includes an encoding module, a temporal distillation module, a cross-channel interaction module, and a regression prediction module. The encoding module includes three parallel color spectral encoders for red, blue, and green. This invention effectively overcomes the technical limitations of traditional air quality prediction methods based on RGB images, such as insufficient temporal resolution, limited dynamic information expression capabilities, and insufficient utilization of information from a single data source. It achieves high-precision estimation of air particulate matter concentration and maintains good robustness and adaptability even under complex environmental conditions.

This invention provides an intelligent communication retransmission method, apparatus, electronic device, and storage medium, applied at a transmitting end, comprising: acquiring multi-stream semantic feature data corresponding to first modal data and second modal data; acquiring data frames through the multi-stream semantic feature data, and acquiring semantic channel reference signals through common semantic features in the multi-stream semantic feature data; transmitting the semantic channel reference signals and data frames to a receiving end, so that the receiving end inputs the data frames and semantic channel reference signals to a semantic verification decoder for decoding, and transmits an indication signal output by the semantic verification decoder to the transmitting end for indicating the channel state, the indication signal including an ACK signal and a NACK signal; and performing a retransmission operation upon receiving a NACK signal. This solves the problem in the prior art where the determination of whether retransmission is needed is only made after all transmitted data has been reconstructed, leading to resource waste and transmission delays.

The invention provides a multi-stream task scheduling method and device of a hardware accelerator, equipment and a storage medium, and relates to the technical field of electric digital data processing. The method comprises the steps of determining a plurality of streams based on a target task of a hardware accelerator, and obtaining an ID and a priority of each stream; all the streams are stored in a red-black tree storage structure based on the ID and the priority of each stream; judging whether an event dependency relationship exists between the streams or not; and if the event dependency relationship exists between the streams, determining an execution sequence of the plurality of streams based on the IDs and priorities of the two streams having the event dependency relationship, and executing the target task of the hardware accelerator according to the execution sequence. According to the method, the flow management and scheduling efficiency can be improved, the key task flow can be quickly executed, delay is reduced, the situation that the high-priority task flow waits for the low-priority task flow is avoided, and the throughput of overall task execution is improved.

The application discloses a high-efficiency GPU multi-stream model inference operator scheduling system and method, and the system comprises the following modules: a global operator scheduler module, a real-time resource sensing module and an operator information database module; the application is directed to a concurrent multi-stream DNN inference service system, and through fusing GPU resource sensing, operator resource demand analysis and a real-time queue scheduling mechanism, dynamic and efficient scheduling of GPU multi-stream operators is realized; the application significantly improves GPU resource utilization, reduces average inference delay in a multi-stream concurrent scene, enhances the stability and throughput capacity of the system under high-concurrent load, and has strong universality and scalability.

The application provides a multi-flow task scheduling method and device of a hardware accelerator, equipment and a storage medium, and relates to the technical field of electric digital data processing. The method comprises the following steps: determining a plurality of flows based on a target task of the hardware accelerator, and obtaining an ID and a priority of each flow; storing all the flows into a red-black tree storage structure based on the ID and the priority of each flow; judging whether there is an event dependency relationship between the flows; if there is an event dependency relationship between the flows, determining an execution order of the plurality of flows based on the ID and the priority of the two flows with the event dependency relationship, and executing the target task of the hardware accelerator according to the execution order. The application can improve the flow management and scheduling efficiency, ensure that the key task flow can be quickly executed to reduce the delay, avoid the high-priority task flow waiting for the low-priority task flow, and improve the overall execution throughput of the task.

The invention provides a multi-stream signal transmission method of a lightning docking station and the lightning docking station, and solves the problems of low bandwidth and insufficient high-speed stability and compatibility of the existing lightning docking station. The method comprises the following steps: acquiring a link total bandwidth, equipment capability, link signal processing configuration parameters and a host end serial differential signal; a synchronous baseband digital serial signal is obtained through preprocessing; obtaining multiple paths of independent service transport streams and corresponding type identifiers through protocol de-framing and service analysis; determining the real-time bandwidth requirement of each stream in combination with the identifier, the link and the equipment parameters; differential signal coding optimization and global time sequence synchronous calibration and scheduling are completed, and multiple paths of differential signals are output to corresponding downlink interfaces; and synchronously acquiring link real-time state parameters, generating adaptive optimization parameters, dynamically adjusting link signal processing configuration parameters, and performing signal transmission with the terminal equipment according to the parameters. According to the scheme, high-speed transmission of data is realized, and the stability and compatibility of the lightning docking station are improved.

A count of multi-stream restore (MSR) streams associated with restore jobs pinned for the MSR streams and a count of MSR streams associated with restore jobs not pinned for the MSR streams are tracked. Each MSR stream includes multiple internal threads that prefetch data for the restore jobs into read-ahead caches. A REST request is received for a restore job, the REST request requesting that the restore job be pinned for an MSR stream. When the count of MSR streams associated with restore jobs pinned for the MSR streams plus the count of MSR streams associated with restore jobs not pinned for MSR streams has reached a threshold, an MSR stream associated with a restore job not pinned for the MSR stream is torn down.

Methods, systems, and apparatus, including computer programs encoded on computer storage media related to multi-stream video encoding for screen sharing a communications session. The system may determine an active pixel area and a remaining pixel area of a video region. of a video region. A first video stream of the active pixel area is generated at a first frame rate. A second video stream of the remaining pixel area is generated at second frame rate, where the second frame rate is a frame rate lower than the first frame rate. A client device may transmit the first video stream and the second video stream to a second client device.

The invention discloses a video frame insertion reconstruction method and system based on spatio-temporal feature fusion. The method comprises the following steps: dividing a video data set into a training set and a test set; feature recovery is carried out on the complex regions in the time dimension and the space dimension to obtain reconstructed features; estimating a plurality of fine-grained optical streams through cascaded multi-stream estimation units in a coarse-to-fine strategy; guiding a pixel warping process through a plurality of attention weight maps of the optical flow warping unit; weighting and mixing all the predicted distortion frames through the mixed mask to generate a reconstructed target frame; and carrying out hierarchical stacking on the multi-stream estimation unit, the attention weight map and the mixed mask to obtain a spatio-temporal feature fused video frame insertion reconstruction model. According to the method, modeling is carried out on target features through two layers of time and space, so that fine textures of the regions are recovered, and the processing capability of the model on complex textures is enhanced.

Embodiments of this application disclose example data transmission method; and example communication apparatuses. One example method includes receiving, by a first core network function entity, multi-stream transmission information of a service, where the multi-stream transmission information indicates that the service is transmitted via first split data and second split data. The first core network function entity can then receive, based on the multi-stream transmission information, first quality of service QoS configuration information and second QoS configuration information, where the first QoS configuration information corresponds to the first split data, and the second QoS configuration information corresponds to the second split data. The first core network function entity can then output the first QoS configuration information and the second QoS configuration information.

The application provides a multi-flow signal transmission method of a thunderbolt expansion dock and the thunderbolt expansion dock, and solves the problems of low bandwidth, insufficient high-speed stability and compatibility of the existing thunderbolt expansion dock. The method comprises the following steps: obtaining a total link bandwidth, device capacity, link signal processing configuration parameters and a host end serial differential signal; obtaining a synchronous baseband digital serial signal through preprocessing; obtaining a plurality of independent service transmission flows and corresponding type identifiers through protocol frame resolution and service analysis; determining real-time bandwidth requirements of each flow in combination with the identifiers, link and device parameters; completing differential signal encoding optimization, global timing synchronization calibration and scheduling, and outputting a plurality of differential signals to corresponding downstream interfaces; synchronously collecting real-time state parameters of the link, generating adaptive optimization parameters, dynamically adjusting the link signal processing configuration parameters, and performing signal transmission according to the parameters and terminal devices. The scheme realizes high-speed transmission of data, and improves the stability and compatibility of the thunderbolt expansion dock.

A network interface device for implementing multi-stream scheduling for time sensitive networking includes direct memory access (DMA) circuitry, comprising: descriptor parsing circuitry to read a packet descriptor from a descriptor cache, wherein the packet descriptor includes at least one scheduling control parameter including: a launch time offset, a gate cycle offset, or a reduction ratio; wherein the packet descriptor is associated with a packet stream having a traffic class; and scheduling circuitry to schedule packets from the packet stream for transmission using the at least one scheduling control parameter.

A multi-flow-path process gas online analysis and pretreatment device. Two-position three-way valves (102, 202, 302, 402, 502, 602) are used for switching flow directions of gases, and switching between communication of a total exhaust pipe (01) and gas inlet paths (1, 2, 3, 4, 5, 6) and communication of a total sample detection gas path and the gas inlet paths (1, 2, 3, 4, 5, 6); according to different requirements, a multi-way switching valve (02) is used for switching between different gas inlet paths (1, 2, 3, 4, 5, 6) and the total sample detection gas path for communication; a first-stage pressure reducing valve (03) is used for regulating the pressure of sample gases to be detected and reduce the pressure of a gas having a pressure higher than a preset value; an impurity filter (04) is used for removing mechanical impurities from the gases to be detected; a second-stage pressure reducing valve (05) is used for performing secondary pressure regulation on the gas decompressed by the first-stage pressure reducing valve (03); and a three-way filter (06) is used for removing mechanical impurities, water vapor and oil hydrocarbons from the gases to be detected, and preventing the impurities from entering a key apparatus or element. The multi-flow-path process gas online analysis and pretreatment device effectively improves the timeliness of monitoring data, reduces errors generated in manual sampling, transportation and analysis processes, and achieves timely and reliable data analysis, thereby reducing the gas analysis cost in processes, and improving the efficiency.

Disclosed is a method of operating a storage device including a plurality of chips, the method including: receiving a write request of stream data; checking a program status or a stream queue of each of the plurality of chips; based on a determination that there is at least one chip in a ready state with an empty stream queue among the plurality of chips, allocating the stream data to the at least one chip in the ready state; and based on a determination that there is no chip in the ready state with the empty stream queue among the plurality of chips, allocating the stream data to at least one chip based on a number of streams in a stream queue among the plurality of chips.

A video recording system and a method for tagging network meetings and video / audio clips from a main video / audio stream are described and claimed. The video recording system allows a system user to view and selectively tag a main video stream or video clips. A rewind control plays back the main video stream and then tags the selected video clip from a start indicator to an end indicator. The video recording system also allows network members to conduct teleconferences while a main video stream is being recorded. Any network member can tag selected sections of the teleconference in real time. In addition, a recording alert control automatically displays a reminder message to the operator of a digital device after a period of inactivity, such as 5 to 10 seconds, following the opening of the software application.This feature is useful when the camera operator forgets to record. A voice control function allows you to activate the controls using voice commands.

The present invention provides a method, apparatus, and device for configuring multi-stream bandwidth for 400GE Ethernet. The method comprises: obtaining at least two data streams to be encapsulated; determining the number of data packets transmitted by each of the at least two data streams within a target maximum transmission period; determining, based on the number of data packets transmitted by each of the at least two data streams, a packet gap register parameter and the number of configuration table entries for each of the at least two data streams; obtaining a configured bit rate for each data stream based on the packet gap register parameter and the number of configuration table entries for each of the at least two data streams; and implementing bandwidth configuration for the at least two data streams based on the configured bit rate. The solution of the present invention can achieve precise bandwidth percentage allocation for 400G Ethernet networks.

The invention relates to a test tool system for a U6G frequency band multi-stream large-scale antenna array, a BOT layer of a substrate is provided with a plurality of channel array groups which are arranged in an array and are mutually isolated, and each channel array group comprises a plurality of channel units arranged in an array; each channel unit comprises a plurality of one-to-two power dividers which are cascaded in a symmetrical mode, the channel units are connected in series end to end through a first transmission line and a second transmission line, and signal interference is isolated through a grounding shielding wall; the output ends of the one-to-two power dividers are connected through a first phase difference line and a second phase difference line; the first phase difference line is a reference phase, and the second phase difference line is a compensation phase. The one-to-two power dividers are cascaded in a tree-shaped symmetrical mode, phase errors of transmission lines are offset, and then the two differential lines with different phases are connected to compensate phase deviation caused by machining tolerance; and a grounding shielding wall is arranged at the periphery of the channel unit so as to isolate signal interference between the channel units.

Embodiments of the present invention disclose a malicious traffic detection method, device, and electronic device, relating to the field of computer network security technology. The method comprises the following steps: obtaining a data stream to be detected belonging to the same network address; splitting the data stream to be detected into multiple TCP single streams; extracting single-stream features of each TCP single stream; generating multi-stream features of multiple TCP single streams based on the single-stream features of the multiple TCP single streams as flow features of the data stream to be detected, wherein the multi-stream features represent the correlation between the multiple TCP single streams; inputting the flow features of the data stream to be detected into a pre-trained malicious traffic detection model to obtain a detection result indicating the malicious threat level of the data stream to be detected, wherein the malicious traffic detection model is trained using a machine learning model based on multiple historical data streams with known malicious threat levels; and determining whether the data stream to be detected is a preset type of malicious traffic based on the detection result. This solution improves the efficiency and accuracy of malicious traffic detection.

The application provides a multi-flow fusion gene sequence prediction method based on Mamba and a double coding strategy, aims to improve the semantic understanding and modeling ability of DNA sequences, and thus improves the accuracy and generalization performance of protein expression value prediction. The method synchronously processes the forward chain and the reverse complementary chain of the DNA sequence: the reverse complementary chain is segmented and vectorized by BPE coding to capture semantic fragments; the forward chain is encoded by one-hot and linearly mapped to a dense representation. After extracting deep features by Mamba Block, the two features are spliced and fused, and then input into LSTM to model the context dependence, and finally output the expression value through the full connection layer. The method effectively solves the problem that One-Hot coding is difficult to capture biological semantic fragments and BPE coding may split the functional domain. Based on the Mamba architecture, the local structure and global semantics are effectively combined, the long-distance dependence is captured while keeping low computational complexity, and the method is suitable for tasks such as gene design and protein expression regulation.

Improving read performance using pre-fetches by encoding an extent map provided by an application executed on a client, where the extent map identifies valid data of the file as extents that may be interspersed with holes constituting invalid data. A server-based filesystem interprets the extent map as a read receipt to bring the right extents in memory ahead of a read request. The extents of the extent map are prefetched at a prefetch read horizon. Because the client is aware of the application-resident extent map, the prefetch operations can intelligently skip over the holes. Multi-stream restores using multiple streams to issue read-ahead operations for the pre-fetches in parallel can also be accommodated.

A method of providing a time-synchronized multi-stream data transmission includes the steps of: (i) providing a first recording or generating device for providing frames of a first data stream; (ii) encoding the first data stream into packets, wherein each first data stream packet includes a presentation time and a predetermined number of first data stream frames; (iii) providing a second recording or generating device for providing frames of a second data stream; (iv) encoding the second data stream into packets, wherein each second data stream packet includes a presentation time and a predetermined number of second data stream frames; (v) transmitting the first and second data streams to a processing device; (vi) associating a start reference time with the first packet of the first and second data streams transmitted; (vii) assigning a first stream start time to the first packet of the first data stream transmitted and a second stream start time to the first packet of the second data stream transmitted; (viii) associating a first time offset from the start reference time and first data stream start time (either directly or indirectly) with each subsequent packet of the first data stream and associating a second time offset from the start reference time and second data stream start time (either directly or indirectly) with each subsequent packet of the second data stream; (ix) time-aligning the first data stream packets with the second data stream packets using the first time offset associated with the first data stream and using the second time offset associated with the second data stream; and (x) applying a common presentation time to the aligned first and second data stream packets, irrespective of the original presentation time of the first and second data streams.

This invention discloses a salient object detection method based on a multi-stream feedback network framework, comprising: acquiring the original image to be detected; inputting the original image into a trained multi-stream feedback saliency prediction model to obtain a saliency prediction result; wherein the prediction model includes a prediction network with shared parameters across multiple streams, and the prediction method is as follows: except for the first-stream prediction network, each subsequent stream prediction network enhances the saliency features of the original image based on the prediction map of the previous stream prediction network to obtain an enhanced image; after feature extraction, aggregation, convolution, and upsampling processing of the enhanced image, prediction maps of each stream prediction network are obtained; the prediction map of the last stream is used as the final saliency prediction result. This method utilizes the features of the enhanced original image from the prediction maps of previous streams as input to the current stream prediction network, and the parameters of each stream network are shared, thus improving the accuracy of salient object detection while maintaining high detection efficiency.

The present disclosure provides methods, components, devices, and systems for a beam search procedure for multiple spatial streams. Some aspects more specifically relate to a beamforming training procedure for multiple spatial streams. The beamforming training procedure can provide beam separation between beams during the training procedure to prevent selection of beams that result in inter-stream interference. In some examples, the multi-stream beamforming training procedure described herein can provide beam separation using antenna polarization. For example, different radio frequency (RF) chains can use different antenna polarizations. Additionally or alternatively, the multi-stream beamforming training procedure can provide beam separation by antenna orientation, with different RF chains using mutually exclusive beam search regions. Additionally or alternatively, a responding wireless communication device can select beamforming directions for spatial streams iteratively or one at a time, and an initiating wireless communication device can perform an additional beam search excluding directions within a beam width of previously selected beams.

The invention relates to computing, and more particularly to (sub)processors in an associative, synchronous, FIFO register, conversion memory. The technical results consist in an increase in reliable diagnostic capability in SIMD mode, and conflict-free multi-stream processing of instructions and data in MIMD mode. A configurable associative, FIFO register, conversion memory is configured in the form of a computing surface comprised of operating bits and permits the formation of a part or the whole of a connected orientable computing surface of genus 0 or more. The claimed method comprises (micro)programmably configuring or reconfiguring a computing surface by a part or the whole of a connected orientable surface of genus 0 or more, carrying out testing in SIMD mode, excluding operating bits that fail information transfer and / or conversion process tests, (micro)programmably changing the genus of the connected orientable computing surface, and using said surface in information processes.