194 results about "Weight prediction" patented technology

Provided are image encoding and decoding methods and apparatuses using weighted bi-prediction. The image decoding method and apparatus according to one embodiment obtain one or more prefixes including a first counter value indicating the number of additional syntax elements subsequent to main syntax elements from a received bitstream, obtain the additional syntax elements including weight information and information about whether to perform weighted bi-prediction on prediction units separated for predicting a current block, based on the one or more prefixes, and generate a prediction block including a prediction value, based on the main syntax elements and the additional syntax elements.

A method and apparatus effectively encode multiple layers using interlayer information in a multi-layered video codec. The method includes reading information concerning the low layer picture, calculating weighting factors using the information concerning the low layer picture, calculating a weighted sum of reference pictures for the current picture using the weighting factors and generating a predicted picture for the current picture, and encoding a difference between the current picture and the predicted picture.

The invention relates to a cow shape measuring method and a cow shape measuring system based on a stereo vision technology. The method comprises the following steps: collecting image information of cow samples at different angles, and calculating body size data of the cow samples according to the image information at different angles; acquiring weight data of the cow samples, and building a weight prediction model according to the corresponding relationship between body size data and weight data; calculating body size data of a cow to be measured according to the image information of the cow to be measured at different angles, and inputting the body size data of the cow to be measured into the weight prediction model to estimate the weight data of the cow to be measured. By using the technical scheme provided by the embodiment of the invention, the problem that weight data measured by a weighbridge is inaccurate due to dynamic nature of cows is solved; and the body size data of cows is acquired through image acquisition, so that the problem that an artificial field contact method is time-consuming and labor-consuming, is of strong subjectivity, and causes stress reaction of cows is solved, and full-automatic cow body size data measurement and weight data measurement is realized.

The invention relates to a fish body dimension and weight rapid acquisition apparatus. The apparatus comprises a window-shaped fixing device and a dimension reference object. The upper inner wall of the window-shaped fixing device is provided with a high-angle shooting pick-up head. One side wall of the window-shaped fixing device is provided with a lateral-shooting pick-up head. One side wall opposed to the side wall where the lateral-shooting pick-up head is located is provided with an LED lateral-shooting lamp box. The lower inner wall of the window-shaped fixing device is provided with an LED high-angle shooting lamp box. The bottom of the window-shaped fixing device is further provided with an electronic balance. The acquisition method of the acquisition apparatus comprises the following steps of shooting a background image respectively in a high-angle shooting way and a lateral shooting way when no fish is present; shooting a dimension reference object image respectively in a high-angle shooting way and a lateral shooting way, and setting up a model for fish body dimension calculation and fish weight prediction. The fish body dimension and weight rapid acquisition apparatus which can rapidly acquire multi-parameter fish body dimension indexes and weights of fish bodies can further set up a fish body dimension and weight prediction model in a computer. The prediction precision can reach more than 90%. The model can be applied to research and development of fish weight sensors.

The invention provides a coding and decoding method and apparatus, and the method comprises the steps: obtaining a weighted prediction value of a current block through the following steps when it is determined that intra-frame and inter-frame weighted prediction is started, and enabling the weighted prediction value to be used for the coding or decoding of the current block; obtaining a first prediction mode of a first adjacent block of the current block and a second prediction mode of a second adjacent block of the current block; determining a target prediction mode of intra-frame predictionaccording to the first prediction mode and the second prediction mode, and obtaining an intra-frame prediction value of the current block according to the target prediction mode; obtaining a motion information candidate list of the current block, wherein the motion information candidate list comprises at least one piece of motion information; determining target motion information of inter-frame prediction according to the motion information candidate list, and obtaining an inter-frame prediction value of the current block according to the target motion information; and carrying out weighting processing on the intra-frame prediction value and the inter-frame prediction value to obtain the weighted prediction value. Through the technical scheme of the invention, the prediction performance can be improved.

The invention relates to a grain storage quantity detection method for horizontal warehouses and shallow silos. The grain storage quantity detection method comprises the following steps of: firstly selecting specific pressure sensors to lay a grain warehouse pressure sensor detection network; distributing the pressure sensors along the side wall of the grain warehouse in the mode that all the pressure sensors are evenly-spaced laid at d meter distance from the side wall; and determining whether a grain warehouse is calibrated or not so as to obtain a calibration parameter and a grain warehouse weight prediction model, spreading the top surface of a grain pile in the grain warehouse after the grain warehouse is fully filled with the grain, detecting the average output valve of the pressure sensor at the bottom of the grain warehouse, and detecting the grain storage quantity of the grain warehouse according to an actually-tested bottom surface area of the grain warehouse, a bottom surface area of the calibrated grain warehouse and calibrated parameters by using a model so as to obtain an estimated value of the weight of the grain warehouse. According to the grain storage quantity detection method provided by the invention, the pressure sensors are required to be arranged on the bottom surface of the grain warehouse, and a compensating item of a side grain warehouse friction action is added into a bottom surface pressure estimating model in order to avoid using the side pressure sensor and accurately obtain the estimated value of the pressure of the bottom surface of the grain warehouse, so that the quantity of the sensors is obviously reduced, and the device detection cost and the installation and maintenance cost are reduced.

The present invention relates to an image processing apparatus and an image processing method capable of performing weighted prediction on the basis of local characteristics of an image.

An inter-TP motion prediction/compensation unit 76 performs a matching process on a block of an image of a frame to be encoded using an inter-template matching method and performs implicit weighted prediction using a weighting coefficient computed from the pixel values of a template region for the matching. The weighting coefficient is computed by a weighting coefficient computing unit 77. The present invention is applicable to, for example, an image encoding apparatus that performs encoding using the H.264/AVC standard.

A decoder (10) conceals errors in a coded image comprised of a stream of macroblocks by examining each macroblock for pixel errors. If such errors exist, then each of at least two macroblocks pictures from each of two different pictures are weighted to yield a weighted prediction (WP) for estimating missing/corrupt values to conceal the macroblock found to have pixel errors.

The invention relates to a text message based waveform concatenation speech synthesis method. The text message based waveform concatenation speech synthesis method includes steps of S1, extracting acoustic parameters and text parameters of all elements in an original voice frequency through segment cutting, and training a duration prediction model and a weight prediction model according to extracted parameters; S2, using a layered pre-selection method to primarily pre-select the elements in a corpus to obtain candidate elements by means of a target element of text analysis and a duration predicted by the duration prediction model; S3, calculating the target element, the candidate elements, and weight information predicted by the weight prediction model to obtain a target cost; calculating Integrating degrees of two adjacent elements to obtain a concatenation cost; using a viterbi searching method to search the target cost and the concatenation cost to obtain a minimum cost path so as to further obtain an optimum element and obtain synthesis speeches through smooth concatenation.

The invention belongs to the field of video coding and decoding technology, and discloses a method for making an inter-frame weighted prediction based on a reference quality. The method comprises thefollowing steps: for a coding mode that requires motion estimation, first calculating a weight of a corresponding reference list based on a reference frame quality to obtain a first weight mode; thenselecting a preferred weight mode from the first weight mode and a default weight mode in combination with a rate distortion cost, searching an optimal matching block for a current coded block, and performing a weighted prediction in the preferred weight mode; and for a coding mode that does not require motion estimation, selecting an optimal weight from the weights of coded blocks at all candidate list positions of the current coded block based on the rate distortion cost, and then performing a weighted prediction based on the optimal weight. The method not only can improve the accuracy of searching the matching block, but also can reflect a different matching degree to the current coded block when a motion compensation block performs weighting. The calculation of weights is adaptive, which reduces the complexity.

The invention relates to an undirected network edge-connectivity weight prediction method based on node similarity. The undirected network edge-connectivity weight prediction method comprises following steps of 1) establishing an undirected network diagram in dependence on an existing data set including nodes and the edge-connectivity weight; 2) respectively calculating following three kinds of similarity indexes including the local similarity index, the global similarity index and the semi-local similarity index in the network diagram in step 1); and 3) predicting the edge-connectivity weight in the test set in dependence on the three kinds of similarity indexes calculated in the step 2) and by means of a multivariate linear regression model, and then examining the model mean performances including the Pearson coefficient and the root-mean-square value by means of a tenfold crossing verification method. According to the invention, by means of the node similarity, the missing edge-connectivity weight is predicted through the multivariate linear regression model, the model is simple, and the prediction result is good.

Systems and methods for determining body weight predictions and human conditions are disclosed. A body weight may be predicted by capturing at least one image of a human, and determining, from the image, a body weight prediction of the human by processing the at least one image with a data processor. The body weight prediction may further be based on an age-based weight factor. A model such as a neural network model may be used to predict body weight.

An image coding apparatus which compression-codes moving picture data including a current image to be coded, the image coding apparatus including an accumulating unit which accumulates a reference image, a predicted image generating unit which generates a predicted image of the current image to be coded, based on the reference image accumulated in the accumulating unit, by performing motion compensation with weighted prediction or motion compensation without the weighted prediction, the weighted prediction being a process of adding a predetermined offset value to a pixel value of the reference image, a coding unit which codes a difference between the predicted image generated by the predicted image generating unit and the current image to be coded, and a determining unit which determines whether or not the current image to be coded is included in, among the moving picture data, a picture group which is a group of successive pictures and indicates motion equal to or larger than a predetermined amount of motion, in which the predicted image generating unit generates the predicted image by performing the motion compensation without the weighted prediction when the determining unit determines that the current image to be coded is included in the picture group, and to generate the predicted image by performing the motion compensation with the weighted prediction when it is determined that the current image to be coded is not included in the picture group.

The invention discloses a speech dereverberation method based on depth characteristics of a generative adversarial network, the method mainly comprises the following steps of: carrying out signal preprocessing on speech through weighted prediction error (WPE); carrying out characteristic extraction of data: carrying out MFCC characteristic extraction on speech data and bottleneck characteristic BNF extraction aiming at phonemes without reverberation speech; then carrying out signal preprocessing on speech through WPE: constructing the generative adversarial network with depth characteristics containing the MFCC characteristics of reverberation speech mapped to clean speech through the generative adversarial network; finally carrying out forced alignment by using a traditional GMM-HMM through a Kaldi toolbox, and then carrying out acoustic model training and decoding by using a depth neural network. The speech dereverberation method combines a method of signal processing with a deep learning framework based on the generative adversarial network to enable a system to combine the respective advantages of both to produce a better speech dereverberation effect.

Embodiments of the invention provide a method and a device for measuring the weight of a live pig, and relate to the technical field of artificial intelligence. The method comprises the following steps: acquiring a video image about the to-be-measured live pig; inputting the video image into a pre-trained key point detection model to obtain a key point thermodynamic diagram and key point positioninformation of the to-be-measured live pig output by the model; obtaining a plurality of preset key points according to the key point position information; calculating the hip width, the hip height and the body length of the to-be-measured live pig according to the preset key points of the to-be-measured live pig; calculating the size of the to-be-measured live pig according to the hip width, thehip height and the body length; and inputting the hip width, the hip height, the body length and the size into a preset live pig weight regression model for calculation to obtain a weight prediction value of the to-be-measured live pig. According to the technical scheme, the problem of low accuracy of visually measuring the weight of the live pig in the prior art can be solved.

The invention provides a least square support vector machine (LSSVM) periodic weighting prediction method based on wavelet and chaos optimization. The method utilizes the wavelet decomposition technology to enable selected sample set data to be decomposed into components of different frequencies. A component phase space is reconstructed on the basis of the chaos theory, the reconstructed components are trained by using LSSVM models, parameters of LSSVM prediction models are optimized through chaos particle swarm optimization, and finally wavelet reconstruction is performed on prediction components obtained by the LSSVM models to whole periodic weighting load prediction waveforms. Through wavelet decomposition of a periodic load, reconstruction of a load phase space on the basis of the chaos theory, construction of a least square support vector machine, and chaos particle swarm optimization of the parameters, the LSSVM periodic weighting prediction method based on the wavelet and the chaos optimization can predict periodic weighting load waves at some period of reconstructed waves when load time sequence has certain chaos characteristics and can be widely used for prediction of periodic weighting load to be born of a support.

The invention relates to a power grid security risk prediction method. The method comprises steps that screened network security posture factors are processed; security posture state classification standards are determined; a hidden Markov model is constructed according to actual power grid conditions; hidden Markov model parameters are updated according to present security posture of equipment; security posture of an integral electric power network is calculated; a prediction risk value of the equipment and a weight prediction risk value are calculated. Through the method, the network security posture factors are deeply researched, the factors having greatest influence on the security posture are screened out, data processing workload is reduced, model training is carried out through utilizing the data relevant to the security posture factors, a problem of over-hugeness of a hidden Markov transition matrix is effectively alleviated, moreover, problems of non-high prediction precision and parameter determination difficulty are alleviated.

The invention discloses a text information-based waveform concatenation voice synthesizing method. The method comprises the following steps: extracting acoustic parameters and text parameters of all primitives from original audio through sound segment segmentation, and training a time length prediction model and a weight prediction model according to the extracted parameters; performing layered pre-selection by adopting a layered pre-selecting method; performing primary pre-selection on the primitives in a language database by utilizing a target primitive analyzed from a text and the time length predicted by the time length prediction model to obtain candidate primitives; calculating the target primitive, the candidate primitives and the weight information predicted by the weight prediction model to obtain target cost; calculating the integrating degree between two adjacent primitives to obtain concatenation cost; searching the target cost and the concatenation cost by using a Viterbi search method to obtain the least cost path, so as to obtain an optimal primitive and a synthesized voice through smooth concatenation. According to the method, the voice synthesizing efficiency is improved; the real-time performance of the concatenation synthesis is enhanced; the prosodic features of the synthesized voice are improved.

The invention relates to a video weighted prediction system and a coding and decoding method thereof based on the H.264 standard and the AVS standard in the technical field of image processing, comprising an arithmetic module, a control module, a parameter storage module and a lookup table, wherein, the arithmetic module is connected with the control module so as to transmit a weighted prediction parameter, the control module is respectively connected with the arithmetic module, the parameter storage module and the lookup table so as to transmit the weighted prediction parameter and a control instruction, and the arithmetic module receives a weighted prediction request and outputs a weighted prediction result. The invention uses a set of weighted prediction arithmetic circuit, and input data for forward weighted prediction and rearward weighted prediction does not need to be generated simultaneously, thereby meeting the requirement on speed, saving hardware resources and efficiently accomplishing the H.264 standard and the AVS standard weighted prediction.

Techniques are described for video encoding and decoding with same picture or count numbering for scalability support. One example involves obtaining a portion of a picture (e.g., a slice, a block, or other portion) and determining if weighted prediction is enabled for the portion of the picture. When weighted prediction is enabled, a zero value picture order count offset indicating a reference picture from a reference picture can be used, though different portion of the picture (e.g., different slices of a picture) can have different picture order count offset values. The portion of the picture can then be reconstructed using the reference picture identified by the zero value picture order count offset. Additional embodiments can use weighted prediction flags and different offset value determinations to support scalability or reference pictures in weighted prediction with a different size than the picture being reconstructed.

A method of decoding a video signal according to the present invention may include: determining whether or not change in brightness is present between a current video including a current block and a reference video of the current video; when it is determined that change in brightness is present between the current video and the reference video, determining weight prediction parameter candidates of the current block; determining a weight prediction parameter of the current block on the basis of index information specifying any of the weight prediction parameter candidates; and performing prediction for the current block on the basis of the weight prediction parameter.