620results about How to "Improve encoding performance" patented technology

The invention includes a medical document handling system and method and automated coding systems and methods for assigning predetermined medical codes to medical documents based on the documents' contents. The invention functions by analyzing electronic medical records and extracting medical information using natural language processing and machine learning. The system collects and amalgamates medical documentation in various formats from multiple sources and locations, normalizes the information, analyzes the information, recognizes information indicating contents corresponding to classification codes, assigns classification codes, and presents information in context correlated to medical records for billing and other purposes.

A method is provided for transmitting signals from a transmitter comprising two or more antennas in a mobile telecommunications network. The method involves determining channel state information, estimating the reliability of that channel state information, and space time block encoding at least one data sequence. Before transmitting the data sequence, a linear transformation is applied to the data sequence so as to at least partially compensate for channel variations. The linear transformation is dependent upon the channel state information and upon the estimated reliability of the channel state information.

A scalable video encoding method and apparatus based on a plurality of layers are provided. The video encoding method for encoding a video sequence having a plurality of layers includes coding a residual of a first block existing in a first layer among the plurality of layers; recording the coded residual of the first block on a non-discardable region of a bitstream, if a second block is coded using the first block, the second block existing in a second layer among the plurality of layers and corresponding to the first block; and recording the coded residual of the first block on a discardable region of the bitstream, if a second block is coded without using the first block.

The invention relates to a video encoder (201) for encoding a video signal. The video encoder comprises a segmentation processor (207) which divides the picture into picture regions. Preferably, picture regions having a high degree of flatness or uniformity are determined in this way. A characteristics processor (209) determine a spatial frequency characteristic for each picture region, and a coding controller (211) selects an encoding block size, such as a prediction block size for motion estimation, in response to the spatial frequency characteristic. An encode processor (213) encodes the picture using the selected encoding block size. Specifically, increasing block sizes are selected for increasing degrees of uniformity or flatness indicated by the spatial frequency characteristic. Thereby, an increasing proportion of high frequency components and a consistent choice of encoding block sizes are maintained, and thus the coding artefacts from many encoders having variable prediction block sizes is reduced. The invention is particularly suitable for H.264 and similar encoders.

A method is provided for transmitting signals from a transmitter comprising two or more antennas in a mobile telecommunications network. The method involves determining channel state information, estimating the reliability of that channel state information, and space time block encoding at least one data sequence. Before transmitting the data sequence, a linear transformation is applied to the data sequence so as to at least partially compensate for channel variations. The linear transformation is dependent upon the channel state information and upon the estimated reliability of the channel state information.

The disclosure relates to a method, apparatus, and recording medium for encoding an intra-prediction mode of an image using variable length codes, including: determining an intra prediction mode of a current block; determining an intra prediction mode having smallest mode number among intra prediction modes of encoded adjacent blocks as Most Probable Mode; generating and allocating variable length codes in ascending order from shorter code with higher similarity for directivity with MPM for all intra-prediction modes; and encoding intra prediction mode of current block based on allocated variable length codes, and thereby generates efficient variable length codes in consideration of high correlation in the intra-prediction modes of current block and peripheral blocks even if intra-prediction mode of current block does not match MPM derived with reference to peripheral blocks, and efficiently compressing information on intra-prediction mode of current block using variable length codes, thereby improving intra video encoding performance.

Improved LDPC code structure and concatenation rules for IEEE 802.11n systems, providing two code sets per rate, one longer codeword and one shorter codeword. The longer codeword length is determined by the system parameters, while the shorter codeword length is ⅓ of the longer codeword length. A LDPC concatenation rule is provided, wherein a maximum of one extra OFDM symbol is added in padding in order to reduce the code set granularity. In order to provide improved code performance, more bits are shortened compared with puncturing, by transmitting the extra OFDM symbol. Further, all the required puncturing and shortening bits are distributed across all codewords. This scheme provides a preferred tradeoff between code complexity and over-the-air efficiency.

The invention provides a screen image compression device and a compression method thereof. A plurality of sets of complemental coding tool combinations are used by the compression device, a coding unit with a certain pixel is taken as a coding unit, to a same coding unit, a plurality of sets of coding tool combinations are used simultaneously to carry out precoding, and a combination with an optimal rate-distortion characteristic is selected to compress the coding unit. Simultaneously, the invention provides a decompression apparatus and a decompression method which are corresponding to the compression device and the compressing method, through analyzing input compression code stream data, a decoding tool combination corresponding to the data is selected to decode the data and carry out image reconstruction, and a screen image compression effect of the present invention is raised substantially compared with the prior art.

An inter-frame prediction method in a hybrid video coding standard belongs to the field of video coding. In order to effectively process deformation movement existing in a video sequence, the present invention puts forward an inter-frame prediction method in a hybrid video coding standard for further improving coding performance of a video. The inter-frame prediction method comprises the steps of: obtaining motion information of a plurality of adjacent coded blocks around a current coding block; obtaining a reference index of each dividing unit in the current coding block according to obtained reference indexes of the adjacent coded blocked; and processing motion vectors of the adjacent coded blocks according to the obtained reference indexes of the adjacent coded blocks and the obtained reference index of each dividing unit in the current coding block, so as to obtain a motion vector of each dividing unit in the current coding block. According to the inter-frame prediction method of the present invention, motion information of the current coding block is predicted through the motion information of the adjacent coded blocks of the current coding block, so that deformation movement existing in the video sequence can be effectively described, and coding efficiency can be further improved.

A method for implementing B-frame prediction in video compression comprising the steps of (A) setting a prediction flag (i) “off” if B-frames are used for block prediction and (ii) “on” if B-frames are not used for block prediction, (B) if the prediction flag is off, generating an output video signal in response to an input video signal by performing an inverse quantization step and an inverse transform step in accordance with a predefined coding specification and (C) if the prediction flag is on, bypassing the inverse quantization step and the inverse transform step.

An improved LDPC encoding with concatenation rule and code size with packet lengths specified in octets is provided. The LDPC block size is selected to be an integer number of OFDM tones. The concatenation rule is an improvement of the shortening and puncturing scheme for low data rate. The improvement is two fold. First, the code length is selected based on the number of transmitted information bits in octets (HTLENGTH). Second, it is determined whether one extra OFDM symbol padding should be added based on the number of shortening bits in the last codeword and the available padding bits in the last OFDM symbol.

A method and apparatus for video coding using block partition are disclosed. According to the present invention, if a target block in the current image unit is generated from a first block partition as well as a second block partition, the coding information reuse is applied. According to the coding information reuse, a first set of coding parameters is determined for the target block generated from the first block partition. A second set of coding parameters is determined for the target block generated from the second block partition by reusing at least one encoder coding decision by the target block generated from the second block partition. When different types of blocks are partitioned to generate the same target block, there is no need to independently determine different code parameter sets, thereby reducing the calculation complexity.

The invention discloses a multilayer neural network language model training method and device based on knowledge distillation. The method comprises the steps that firstly, a BERT language model and amulti-layer BILSTM model are constructed to serve as a teacher model and a student model, the constructed BERT language model comprises six layers of transformers, and the multi-layer BILSTM model comprises three layers of BILSTM networks; then, after the text corpus set is preprocessed, the BERT language model is trained to obtain a trained teacher model; and the preprocessed text corpus set is input into a multilayer BILSTM model to train a student model based on a knowledge distillation technology, and different spatial representations are calculated through linear transformation when an embedding layer, a hiding layer and an output layer in a teacher model are learned. Based on the trained student model, the text can be subjected to vector conversion, and then a downstream network is trained to better classify the text. According to the method, the text pre-training efficiency and the accuracy of the text classification task can be effectively improved.

The invention relates to a method and device for performing the movement forecast in the video coding, the method comprises a. selecting a set of reference areas from the coded areas on the current frame; b. estimating the movement vector difference values between the current area and each reference area; c. determining the reference area corresponding to the minimum movement vector difference value, and judging that whether the minimum movement vector difference value is smaller than a preset threshold value; d. setting the movement vector and the reference frame of the current area as the determined movement vector of the reference area and the reference frame, if the minimum movement vector difference value is smaller than the preset threshold value; e. if the minimum movement vector difference value is larger than the preset threshold value, for each forecast direction, setting the movement vector of the current area as 0 vector or an intermediate vector, and setting the reference frame of the current area as the reference frame most adjacent to the current area or the reference frame most constantly adopted by one set of reference areas. According to the invention, the high coding performance can be obtained.

The invention discloses a method and a device for intra-frame prediction and an intra-frame prediction-based encoding / decoding method and an intra-frame prediction-based encoding / decoding system. The prediction method comprises the following steps of: determining adjacent blocks of M*N blocks to be predicted, and selecting corresponding Ki reference sample points from the adjacent blocks according to the length Ki (i is equal to 0, 1, 2,...,(M*N-1)) of a filter at each position i in the blocks to be predicted to form a sample point set for the intra-frame prediction, wherein ki is equal to 0, 1, 2,..., Ki-1, Ki is a natural number and is more than 1, and M and N are natural numbers; selecting a prediction direction m, and determining the prediction weighting coefficient Wi (m, k) of each sample point Ci to be predicted in the blocks to be predicted respectively, wherein i is equal to 0, 1, 2,...,(M*N-1); and calculating prediction values of current sample points to be encoded according to the prediction weighting coefficients Wi (m, k) and the sample point set for the intra-frame prediction. The prediction method improves the prediction precision of an intra-frame prediction mode, so that the integral encoding performance is improved.

Disclosed are an acquisition method and acquisition device for a motion vector, and a video coder and decoder and a method thereof. The acquisition method comprises: acquiring motion information about adjacent blocks of a current block in a current frame and motion information about a first reference block having the same position as the current block in a first reference frame, the motion information comprising a forward motion vector and a forward reference identifier, and a backward motion vector and a backward reference identifier; successively, according to the forward motion vector in each piece of motion information, going to the forward reference frame indicated by the forward reference identifier to determine a forward matching block of the current block, and according to the backward motion vector, going to the backward reference frame indicated by the backward reference identifier to determine a backward matching block of the current block; detecting a similarity between the forward matching block and the backward matching block which are determined each time; and setting the motion information corresponding to the maximum similarity as the motion information about the current block. By implementing the present application, the calculation complexity can be reduced, and the problem in the prior art that the encoding efficiency needs to be further increased can be solved.

A method of data encoding is disclosed. An encoder receives a set of information bits and performs a lifted LDPC encoding operation on the information bits to produce a codeword. The encoder then punctures all lifted bits of the codeword that correspond to one or more punctured base bits of a base LDPC code used for the LDPC encoding operation. The base LDPC code has no multiple edges, and the one or more punctured base bits are those that correspond with one or more punctured base nodes, respectively, of the base LDPC code. For some embodiments, the one or more punctured base nodes correspond to one or more degree 2 variable nodes.

The present invention relates to a video encoding method and apparatus for setting and encoding quantization parameters, and to a video decoding method and apparatus for decoding and setting quantization parameters in a video encoding and decoding apparatus which uses blocks having various sizes and depths as encoding and decoding units.

The invention discloses a method, a device and a system for capturing and coding audio / video of a mobile terminal. The method comprises the following steps: acquiring network type information and self unit type information of the mobile terminal, sending a coding parameter request command carrying the information to a parameter server, analyzing the obtained coding parameter request command by the parameter server, querying a coding parameter database according to the analyzed network type information and self unit type information of the mobile terminal to obtain the corresponding coding parameter, then sending the coding parameter to the corresponding mobile terminal, coding the captured audio / video according to the obtained coding parameter by the mobile terminal, and sending the coded audio / video to an audio / video server. The method, the device and the system allow the mobile terminals with the same unit performance to share the adjusted audio / video coding parameter of the current unit; when an individual mobile terminal requests the coding parameter from the parameter server to code the self audio / video data, a better audio / video coding effect can be achieved.

The invention provides a method for encoding and decoding lattice vector quantization audio. The method comprises the following steps: dividing a frequency domain coefficient of an audio signal which is subjected to modified discrete cosine transform (MDCT) conversion into a plurality of encoding sub-bands, and quantizing and encoding amplitude enveloping value of each encoding sub-band to acquire encoding bit of amplitude envelope; performing bit distribution on each encoding sub-band, normalizing, quantizing and encoding vectors in low-bit encoding sub-bands and high-bit encoding sub-bands by using a tower lattice vector quantizing method and a spherical lattice vector quantizing method to acquire encoding bit of the frequency domain coefficient respectively; and complexly packaging theencoding bit of the amplitude envelope and the encoding bit of the frequency domain coefficient of each encoding sub-band, and transmitting the package to a decoding end. The method and the system for encoding and decoding the lattice vector quantizing audio can acquire better voice source encoding effect.

The invention discloses self-adaptively selecting global motion estimation method for panoramic video coding, which comprises the following steps that: (1) an image is divided into a plurality of sub-image blocks, (2) zero vector block is decided, (3) multi-resolution motion is estimated, (4) translational motion model motion parameter is estimated, (5) six-parameter motion model motion parameter is estimated, (6) twelve-parameter motion model motion parameter is estimated, and (7) RDO optimizes an estimation model of self adaptively choosing global motion. The advantage is that the estimation model of self adaptively choosing global motion optimized by the RDO can well describe the camera motion flow characteristic of the current frame well so that the motion estimation is more accurate and the coding performance is higher. At the same time, aiming at the characteristics of the rich detail and the high resolution ratio of panoramic video motion, the invention takes the Lagrange parameter Lambada=0.85x2QP / 3+2QP / 6 and further obtains a good balance between compression code rate and image distortion degree.

The invention discloses an automatic text classification method based on BERT and feature fusion. The method comprises the following steps: firstly, cleaning text data, realizing conversion from a text to a dynamic word vector through BERT, extracting features of the text by utilizing CNN and BiLSTM, and respectively transmitting a word vector sequence output by the BERT to a CNN network and a BiLSTM network; then, splicing the output of a CNN network and the output of a BiLSTM network together, carrying out feature fusion, and finally, outputting a final prediction probability vector througha full connection layer and a softmax layer. The method is suitable for the general supervised text label prediction problem, and can effectively improve the prediction accuracy of the text data labels with prominent sequence information and local features.

The invention discloses a coding method and a decoding method of point cloud attribute information, a storage medium and terminal equipment. The coding method comprises the following steps: sorting all points in a point cloud according to geometrical information; encoding certain attribute information of the current point and writing the encoded attribute information into a code stream; calculating the difference between the coded attribute information of the current point and the attributes of a plurality of neighbor points; selecting a neighbor with the smallest attribute difference as the prediction reference point of other attribute information of the current point; and performing predictive encoding on other attribute information of the current point and writing into a code stream. According to the invention, the coding result of one piece of attribute information is used for guiding other pieces of attribute information to select the prediction reference point, so that the prediction accuracy is improved, and the coding performance of the point cloud attribute information is further improved.

A method for performing polar encoding includes the steps of, if a length of a codeword is greater than 2x and smaller than 2x+1 determining whether encoding is performed by a polar encoder of size 2x or a polar encoder of size 2x+1 according to whether or not a length of the codeword is smaller than a prescribed value, and if the length of the codeword is smaller than the prescribed value, performing encoding by the polar encoder of the size 2x.

An encoding method and device, mainly composed of: firstly, calculating mapping matrixes corresponding to various views according to camera parameter information corresponding to cameras of various views, the mapping matrix is mapping converting coefficient between 2-dimensional coordinates and 3-dimensional coordinates of each point on image of each view; then obtaining 3-dimensional coordinates of object according to mapping matrix corresponding to current view, 2-dimensional coordinates of object of current view and object depth information, and determining 2-dimensional coordinates of object of other view according to this 3-dimensional coordinates and mapping matrix corresponding to other views; then, calculating relative parallax vector or residual difference, and using this parallax vector or residual difference to carry out encoding operation. Embodiment of this invention leads to higher precision on parallax evaluation among different view with less residual difference, in order to effectively enhance multi-view image encoding effect and simplify image encoding process.

The invention provides a coding and decoding method and equipment thereof, and the method comprises the steps: dividing a current block into a first triangular sub-block and a second triangular sub-block if the feature information of the current block meets a specific condition; obtaining first target motion information of the first triangular sub-block and second target motion information of thesecond triangular sub-block; wherein the first target motion information is different from the second target motion information; and performing encoding processing or decoding processing on the current block according to the first target motion information and the second target motion information. According to the technical scheme, the prediction accuracy can be improved, and the prediction performance can be improved.

The invention discloses an improved HEVC sample point self-adaption compensation method. The BO algorithm and the EO algorithm are adopted for performing SAO filtering on images. The sample point self-adaption compensation algorithm is adopted for solving ring effects caused by transform and quantization in the current video coding standard HEVC in the loop filter technology. The current video coding standard HEVC comprises BO and EO algorithms, but sideband distribution conditions are not fully taken into consideration in the BO algorithm which only transmits four continuous sidebands. According to the invention, by use of an image histogram, the sideband distribution conditions can be sufficiently analyzed; intra-frame prediction mode type texture information is extracted in a self-adaption manner; and for different sizes of CTB, four kinds of BO filter modes are added, BO filter pattern discrimination process is improved, and the improved self-adaption compensation method is provided. The experimental results show that compared with the HM10.0, coding performance is improved; bit rate gain of chromaticity can reach -7.53%; and subjective assessment quality of image edge parts is obviously improved.

The invention discloses a background modeling-based static camera video compression method and a background modeling-based static camera video compression system. The method comprises the following steps: for an input video sequence, selecting a training image set and modeling the selected training image set to generate a background image; adopting a coder to encode and compress the background image to generate a background image code stream; adopting a decoder to decode and reconstruct the background image code stream to acquire a reconstructed background image; differentiating images in the input video sequence and the reconstructed background image, calculating a differentiation result and acquiring a differential image according to the differentiation result; and encoding and compressing the differential image. In the method and the system, a fixed scene part and other parts of video are compressed respectively by utilizing the characteristic that a scene of the video sequence is fixed and establishing and updating a model describing the relatively fixed scene in the video so as to eliminate redundancy in the video sequence to a greater extent to obtain better compression performance.