165 results about "Coding decoding" patented technology

An image encoding / decoding method and device are provided. An image decoding method according to the present disclosure comprises the steps of: determining reference samples separated by a distance of n samples from a current block and reference samples separated by a distance of m samples from the current block (wherein the reference samples separated by the distance of n samples include a top reference sample and a top-right reference sample, and reference samples separated by the distance of m samples include a left reference sample and a bottom-left reference sample); and generating a prediction block of the current block on the basis of a weighted sum of at least two of the determined reference samples, wherein the weight used in the weighted sum calculation is determined on the basis of the n or m, which may be natural numbers.

This invention provides a speech recognition method, system, device, and medium for elevator entrapment scenarios. The method includes: acquiring speech data in an elevator scenario and preprocessing the speech data to obtain a first speech feature; using the first speech feature as input to a deep neural network to output recognized text and an entrapment probability value; determining entrapment based on the recognized text and the entrapment probability value, and outputting an entrapment determination result. By using a deep neural network and a specific encoding and decoding process, this method can more accurately process speech data in elevator scenarios, thereby improving the accuracy of entrapment detection. By comparing the entrapment probability value with a threshold and combining it with the recognized text for comprehensive judgment, this method can reduce false detections and false negatives, improving the accuracy of entrapment determination.

The application provides a low-density parity-check code decoding method and device based on a graph neural network, which comprises the following steps: step 1, constructing a factor graph comprising variable nodes, check nodes and edge relationships, and mapping log-likelihood ratio information of a received signal to initial embedding of the variable nodes; step 2, generating message features according to node embedding, node degree and iteration step length, and constructing attention weights for each edge to measure the importance of the message; step 3, inputting the message features and the attention weights into a gated recurrent unit to update the residual of the edge weight, and feeding back the updated edge weight to the graph structure; step 4, weighting and aggregating the messages from the adjacent nodes according to the edge weight, calculating the updated embedding of the variable nodes and the check nodes, and performing gated modulation combined with the check result; step 5, repeating steps 2 to 4 until a preset iteration number or a decoding convergence condition is reached, and mapping the final variable node embedding to a decoding result to realize LDPC code word recovery. By introducing the attention mechanism and the gated residual update into the message passing process, the application realizes adaptive modeling of the contribution degree of different edge messages, dynamically remembers the historical state, thereby improving the decoding performance and the convergence speed; meanwhile, the application can effectively reduce the bit error rate and is suitable for high-speed reliable data transmission scenarios in a 5G / 6G wireless communication system.

This invention discloses a high-precision phase recovery method for tooth surface fringe images during gear rotation. The method first establishes an N-step static phase-shift image dataset of the gear at subdivided rotational positions. When constructing network training data pairs, three image combinations with different position indices and phases satisfying the standard three-step phase shift are selected to generate an input sequence simulating unidirectional rotation and motion errors. The high-precision sine and cosine terms calculated using the N-step static images are used as labels. Secondly, an encoder-decoder deep neural network with a bottleneck layer dual attention mechanism and global residual connections is constructed and trained using an adaptive loss function. Finally, the dynamic image to be tested is input into the network to directly predict the sine and cosine terms, calculating the wrap-around phase without motion artifacts. This invention effectively suppresses motion errors caused by gear rotation, contributing to the high-precision reconstruction of the three-dimensional morphology of the tooth surface.

The application discloses a source code annotation automatic generation method and system based on word-level retrieval and belongs to the field of natural language processing text generation. A training set composed of code function text, code abstract syntax tree and code annotation text is used to train an encoding-decoding network; an overall feature vector of each annotation word in the code annotation text is obtained to construct a near-neighbor word database; for the code function text to be annotated and the abstract syntax tree thereof, a model-based target word probability distribution and an overall feature vector of a target word at a current time step are generated in a self-recurrent manner; K near-neighbor words with the highest similarity to the overall feature vector of the target word are searched in the near-neighbor word database to generate a target word probability distribution based on the near-neighbor words; and the two target word probability distributions are fused to take the target word with the maximum probability as an annotation word generated at the current time step. The application can greatly improve the annotation generation quality of an original model and also improve the generation probability of low-frequency words in code annotation.

The present invention provides a method and apparatus for encoding / decoding an image to enhance inter prediction. According to the present invention, the method includes: determining a first reference picture and a second reference picture of a current block; determining both a first reference block in the first reference picture and a second reference block in the second reference picture; partitioning the current block into a plurality of sub blocks; and obtaining a prediction block of each of the sub blocks on the basis of reference candidate block index information of each of the sub blocks.

Systems, methods, non-transitory computer-readable media to perform operations associated with the storage medium. One system includes a storage medium and an encoding / decoding (ED) system to perform operations associated with the storage medium, the ED system being configured to process a set of log-likelihood ratios (LLRs) and a syndrome vector to obtain a set of confidence values for each bit of a codeword, estimate an error vector based on selecting one or more bit locations with confidence values from the set of confidence values above threshold value and applying hard decision decoding to the selected one or more bit locations, calculate a sum LLR score for the estimated error vector, and output a decoded codeword based on the estimated error vector and the sum LLR score.

Innovations in range asymmetric numeral systems (“RANS”) encoding and decoding are described herein. Some of these innovations relate to hardware implementations of RANS decoding that organize operations in two stages, which can improve computational efficiency of RANS decoding. Other innovations relate to adapting RANS encoding / decoding to different distributions or patterns of values for symbols. For example, RANS encoding / decoding can be adapted by switching a default symbol width (number of bits per symbol), adjusting the symbol width on a per-fragment basis for different symbol fragments, switching between different static probability models on a per-fragment basis for different symbol fragments, and / or selectively flushing (or preserving) a state of a RANS decoder on a per-fragment basis for different symbol fragments. In many cases, such innovations can improve compression efficiency while also providing computationally efficient performance.

Embodiments of this application provide a method for encoding, a method for decoding, a method for training, and related apparatus. The decoding method may include: acquiring a bitstream D0; determining a reconstructed feature tensor 0 based on the bitstream D0; performing a decoding operation on the reconstructed feature tensor 0 to obtain a reconstructed latent representation 0 using a neural network; and performing a synthesis transform operation based on the reconstructed latent representation 0 to generate a reconstructed signal. Based on the method provided by the embodiments of this application, different features of input data are compressed into different numbers of bits.

An image encoding / decoding method and apparatus for performing bi-directional optical flow (BDOF) and a method for transmitting a bitstream are provided. An image decoding method according to the disclosure is an image decoding method performed by an image decoding apparatus. The image decoding method can include the steps of deriving prediction samples of a current block based on motion information of the current block, determining whether BDOF is to be applied to the current block, if the BDOF is to be applied to the current block, deriving a gradient for a current sub-block in the current block, deriving a refined motion vector (v x ,v y ) for the current sub-block based on the gradient, deriving a BDOF offset based on the gradient and the refined motion vector, and deriving refined prediction samples of the current block based on the prediction samples of the current block and the BDOF offset.

This disclosure provides a method and apparatus for encoding / decoding image signals. The method for decoding image signals according to the present invention may include the following steps: determining whether a brightness change exists between a current image including a current block and a reference image of the current image; if a brightness change is determined to exist between the current image and the reference image, determining candidate weight prediction parameters for the current block; determining weight prediction parameters for the current block based on index information for specifying any one of the candidate weight prediction parameters; and performing prediction for the current block based on the weight prediction parameters.

A video encoding / decoding method and device disclosed herein determine the current block through block division based on a tree structure and sub-sample the current block to form one or more sub-blocks from the current block, and can encode / decode the sub-blocks belonging to the current block according to a prescribed scan order.

The present application relates to the technical field of pavement crack detection, and particularly relates to a lightweight pavement crack detection method, wherein a lightweight feature extraction module SandGlass and a coordinated attention mechanism are integrated into an encoding-decoding structure to construct a lightweight pixel-level crack detection network, a pavement image database is used to train the network model, so that the network model can realize the same detection effect as a larger network model under less computing power, the coordinated attention mechanism module suitable for mobile terminals is introduced to reduce the influence of noise and improve the accuracy of crack detection, and finally the encoding-decoding structure is improved to fuse feature maps of different levels, so that the final prediction result is closer to the actual pavement condition. Further, the improved method can be effectively applied to mobile devices to realize real-time and effective end-to-end detection.

The disclosure provides an image encoding / decoding method, a data transmission method, and a computer readable medium. According to the disclosure, information used in a palette encoding mode can be effectively signaled. For example, information related to availability of the palette mode can be signaled, and information related to a quantization escape value can be signaled based on the information related to availability of the palette mode. In addition, as a quantization parameter of the quantization escape value, minimum quantization parameter information related to a transform skip mode can be signaled. Accordingly, information required for palette mode encoding can be efficiently signaled, and escape encoding efficiency in the palette mode can be improved.

Systems, methods, and apparatus, including computer programs encoded on a computer-readable medium, for encoding, decoding, or transcoding digital video are described. One example method of video processing includes performing a conversion between a video and a bitstream of the video, wherein the bitstream conforms to a sequence of sub-bitstream extraction processes defined by a rule, the rule specifying that (a) the sequence of sub-bitstream extraction processes excludes an operation that removes all supplemental enhancement information (SEI) network abstraction layer (NAL) units that satisfy a condition from an output bitstream, or (b) the sequence of sub-bitstream extraction processes includes an operation that replaces a parameter set with a replacement parameter set prior to removing the SEI NAL units from the output bitstream.

The present invention relates to an image encoding / decoding method and device. The image decoding method according to an embodiment of the present invention may comprise the steps of: deriving context of the current block on the basis of reference samples adjacent to the current block; inputting the context to a neural network model to derive intra prediction-related information about the current block; and generating a prediction block of the current block on the basis of the intra prediction-related information about the current block, wherein the context may include an initial prediction value for the current block generated on the basis of a prescribed intra prediction mode.

This disclosure generally relates to video coding / decoding, and in particular to enhancing block-adaptive weighted prediction. The method involves receiving a coded video bitstream, which includes the current block of the current frame and a first syntax element indicating the prediction mode of the current block, wherein a plurality of scaling factor lookup tables are stored and include different ranges of scaling factors. In multiple scaling factor lookup tables, the step size or precision of the scaling factor in each lookup table is the same, and A step of determining a prediction mode based on the value of a first syntax element, wherein the prediction mode is used to predict the current block based on the reference block of the reference frame. The steps include determining a scaling factor from one of several scaling factor lookup tables, The process includes the step of reconstructing the current block based on a reference block and a determined scaling factor.

The application provides a crowd statistical method and device based on confidence probability, comprising: obtaining an original picture for crowd quantity statistics; obtaining a feature map and a generated map corresponding to the original picture through encoding and decoding based on the original picture; inputting the original picture, the feature map and the generated map into a preset confidence probability model to obtain a confidence probability density map; inputting the feature map into a preset prediction neural network to obtain a prediction probability density map corresponding to the feature map; determining a real density map based on the confidence probability density map and the prediction probability density map, and performing crowd quantity statistics according to the real density map. In one aspect, the application reduces the data training model through encoding and decoding; on the other hand, the application screens the prediction probability density map by using the confidence probability density map, reduces the occurrence of false recognition, and thus improves the accuracy of crowd quantity statistics.

Disclosed herein are a video-decoding method and apparatus and a video encoding method and apparatus, and more particularly a method and an apparatus which perform region-differential image encoding / decoding using a recovered image. In accordance with an encoding method according to an embodiment, a recovered low-quality image is generated by performing encoding on an original image and a recovered high-quality image is generated using the recovered low-quality image. An image is segmented into multiple regions, and encoded reconstruction information for generating a reconstructed high-quality image is generated by performing encoding on the image.

The present invention relates to an image encoding / decoding method and a recording medium therefor. To this end, a method of decoding an image can include the steps of generating a list of a plurality of motion vector candidates according to an inter prediction direction of a current block, obtaining a plurality of motion vectors for the current block by using the list of the plurality of motion vector candidates, determining a plurality of prediction blocks for the current block by using the plurality of motion vectors, and obtaining a final prediction block for the current block based on the plurality of prediction blocks.

This invention relates to a 3D fault segmentation method based on local-global feature enhancement and geometric constraints, belonging to the fields of geophysical exploration and artificial intelligence. The method includes: standardizing and preprocessing the original 3D seismic data volume and cropping it into local sub-blocks; constructing an encoder-decoder network with 3DU-Net as the backbone, where the encoder extracts deep semantic features through 3D convolutional blocks, and the terminal layer aggregates multi-scale geological context information via MC-Bridge; the decoder embeds LG-Module and ACO-Module at each level; training the model using a structured loss function including binary cross-entropy loss, similarity coefficient loss, edge smoothing loss, and geometric constraint loss; and generating a fault prediction probability map by convolution and Softmax activation of the decoded enhanced features. This invention balances local texture details and global long-range dependencies, explicitly constrains fault geometry, and improves segmentation accuracy and geological plausibility.

An image encoding / decoding method and apparatus are provided. The image decoding method includes obtaining, from a bitstream, network abstraction layer (NAL) unit type information of at least one NAL unit including coded image data, determining a NAL unit type of at least one slice in the current picture based on the obtained NAL unit type information, and decoding the at least one slice based on the determined NAL unit type. The current picture includes a plurality of subpictures having different NAL unit types, the bitstream is a single bitstream obtained by merging a plurality of subpicture bitstreams for the subpictures.

An image encoding / decoding method and device are provided. A video decoding method according to the present disclosure comprises the steps of: receiving a bitstream including video information; and generating a reconstructed picture by restoring the current picture on the basis of the image information, wherein the image information may include a post-filter hint SEI message regarding a post-filter to be applied to the reconstructed picture.

The application relates to a bearing fault recognition algorithm suitable for small sample conditions in the field of intelligent manufacturing and industrial fault diagnosis. The whole method process involves vibration signal data acquisition, data cropping, conversion into a frequency domain graph, large-core multi-scale feature extraction, local similarity comparison, global similarity comparison, and fusion of local and global classification stages, and can cope with the real-time performance, high efficiency and deployability of fault recognition under complex working conditions. Among them, the large-core multi-scale feature extraction realizes unified feature extraction modeling; the local module realizes fine-grained feature matching by combining cosine similarity and a multilayer perceptron; the global module models the semantic association between the support set and the query set by using two-dimensional position coding and cross-attention mechanism. Finally, accurate fault classification is realized by fusing the similarity scores of the local and global modules. The method has high accuracy, strong generalization ability and low computational complexity, and is particularly suitable for online diagnosis and edge deployment in industrial scenes.

The application provides an anti-single event upset processing method for avionics, which comprises the following steps: step one, defining and initializing a configurable hierarchical check data structure; step two, entering a loop task and obtaining to-be-written data; step three, writing the to-be-written data and a correction detection mode into the data structure; and step four, when the data needs to be read, reading to-be-read data from the data structure in groups. Compared with a traditional triple modular redundancy scheme, the application reduces storage occupation by more than 60%, solves the problem of high cost of directly using Hamming code for real-time correction and detection of memory data in an aviation device, and reduces resource consumption of encoding and decoding while having correction and detection capability, thereby meeting high requirements of the aviation device on real-time performance and resource efficiency.

The application discloses a terminal positioning method and device, equipment and storage medium, and specifically discloses the following: receiving a ranging code decoding key issued by a core network; receiving encrypted ranging codes broadcast by at least four base stations; using the ranging code decoding key to decrypt the received encrypted ranging codes to obtain original ranging codes corresponding to the base stations; based on the original ranging codes, performing code phase alignment processing on the original ranging codes and copied ranging codes with the same structure as the original ranging codes to obtain the propagation time of signals of the encrypted ranging codes from the base stations to a terminal, and calculating the pseudo distances between the terminal and each base station according to the propagation time; obtaining high-precision geographic positions and clock correction information of the at least four base stations, and based on the pseudo distances corresponding to the at least four base stations respectively, the high-precision geographic positions and the clock correction information, constructing an equation group with the three-dimensional coordinates and clock bias of the terminal as unknowns, and determining the position coordinates of the terminal by solving the equation group.

This invention discloses a one-step snap-fit ​​prediction method for flexible flat cable based on virtual-real fusion and adaptive gradient. First, in the real operating environment of mobile phone flexible flat cable assembly, real tactile images are acquired using a tactile sensor. Then, based on the real tactile images, the Material Point Method (MPM) is used to simulate the real tactile sensation in a virtual environment, generating a virtual tactile image. Real and virtual tactile sensations together constitute a digital twin environment. By performing adaptive gradient calculation on the virtual and real tactile images in the digital twin environment, a one-step prediction model based on adaptive gradients is established. The model uses a CNN+LSTM network to predict the tactile images, generating predicted tactile images, which are then compared with ideal tactile images. Then, by encoding and decoding the tactile images, a one-step snap-fit ​​guidance strategy is obtained and applied in the real operating environment. The one-step prediction model based on adaptive gradients is trained and iterated in the digital twin environment.

Provided are an image encoding / decoding method and device. An image decoding method performed by an image decoding device according to the present disclosure includes the steps of: determining a quantization parameter of the current block on the basis of whether color space conversion is applied to a residual sample of the current block; determining a transform coefficient of the current block on the basis of the quantization parameter; determining the residual sample of the current block by using the transform coefficient; and resetting the value of the residual sample on the basis of whether the color space conversion is applied.