52results about How to "Avoid quantization error" patented technology

The invention discloses a multi-antenna sending method based on a sounding reference signal, a terminal, and a base station. Newly-added SRS sending port information and/or an SRS sending mode for indicating signals sent by a UE standby execution antenna set by turns and a control signal required by the SRS sending mode are sent to UE through eNB. The UE sends the SRS through the SRS sending mode, and the eNB estimates channel information according to the SRS and calculates and sends a pre-coding matrix. By means of the method, the terminal and the base station, the problem that when the number of receiving antennas of the UE is larger than the number of sending ports and the number of downlink configuration ports is larger than 2, downlink pre-codes calculated by the eNB according to the uplink channel matrix can not be matched with wireless channels between all the antennas of the UE and an antenna of the base station is solved; the UE can be used for solving the problem that when the number of sending antennas is larger than the number of the sending ports, the eNB has difficulty in completely obtaining the states of wireless channels between all the antennas of the UE and the eNB and the PUSCH can not be set through the optimal antenna or the optimal antenna set.

The invention belongs to the technical field of cross-modal data retrieval, and particularly relates to a cross-modal hash retrieval algorithm based on a fine-grained similarity matrix. The algorithmprovided by the invention mainly aims at two tasks of image retrieval texts and text retrieval images, and comprises the following steps: hash code reasoning: constructing a fine-grained similarity matrix by utilizing label information of image text pairs, so that hash codes reserve fine-grained similarity information between image text data items; constructing an auto-encoder to enable the hash code to reserve semantic information in the label as much as possible; hash function learning: training two Hash functions, mapping images and texts to Hash codes respectively, wherein target functionsused by Hash code learning include Hash code mapping loss, similarity retention loss with weight and classification loss. The invention has relatively high retrieval precision in two tasks of image search texts and text search images.

In a communication system, a postcoder is coupled between first and second portions of each of a plurality of receivers. The postcoder is configured to receive an adaptive index representation of a scaled signal associated with the first portion of a first one of the receivers, and to process the adaptive index representation using an adjusted postcoder coefficient to generate a postcompensation signal. The postcompensation signal is supplied by the postcoder to the second portion of a second one of the receivers for use in postcompensation of at least one signal received by the second receiver.

The invention relates to an object picture display method with dynamic fusion of a background and a display method thereof. The method comprises the following steps that: (1), color information of an object and transparency information of three primary colors are obtained in advance and are respectively stored as pictures in a JPEG format or a BMP format, and the pictures are stored in a system; (2), a user selected an object that he/she wants to browse by an interface; and (3), a background picture is selected by ways like uploading by a user, hand-operated selection by a user and automatic switching and the like, and the selected background picture is fused with an object picture, so that an effect diagram is shown for the user. In addition, the invention also provides an object picture display system with dynamic fusion of a background. And the system consists of a background processing module, a data storage module, and a foreground display module. According to the invention, the color information and the transparency information of the three primary colors are separated from the object picture with a background and are respectively stored in one picture; an image transparency fusion algorithm is employed to display effect diagrams of the object under different backgrounds, so that the user can fully understand textures and attributes of the object and have an accurate grasp of characteristics of the object.

The invention discloses a computer generated hologram fast generation method for 3D models, comprising: translating and rotating a triangular patch in world coordinate system to standard position, and synchronously doing same transformation on the computer generated hologram, to hold the relative position between the triangular patches in the computer generated hologram and the object face, after the coordinate transformation of the triangular path; deriving Fresnel diffraction integral equation to obtain the relationship between the optical field at one point and the optical field at the top point of the triangular patch in the computer generated hologram; adding the coordinate and optical field of the top point of the special triangular patch into the equation to obtain the optical field at each sampling point on the computer generated hologram. The computer generated hologram fast generation method can significantly reduce calculation amount, avoid sampling error on object face and generate computer generated holograms of various resolutions.

The invention discloses a signal processing method. The method comprises the following steps: a BBU performing bit-slicing processing on a to-be-processed signal after acquiring the to-be-processed signal after gain processing from a RRU, and performing saturation processing on the to-be-processed signal after performing the bit-slicing processing, thereby avoiding quantization error of a signal caused by transmission between boards, meeting the requirement of a system index while reducing data bit width transmitted between the boards, and guaranteeing that the bit-slicing error cannot increase the influence of the quantization error to the sensitivity of a receiver. And meanwhile, the method further comprises the step of adjusting linkage gain according to average power of a signal acquired by an ADC so as to prevent signal from oversaturating in the premise of guaranteeing the precision of the sensitivity of the receiver.

The invention provides the orthogonal propagation operator direction finding method based on a multi-peak quantum cuckoo search mechanism. The method comprises the following steps of through samplingdata, acquiring the estimation values of a covariance matrix and a propagation operator; initializing parameters and quantum bird eggs, and calculating the fitness of the mapped bird eggs; initializing a memory pool; using Levy flight to update the quantum bird eggs and calculating the fitness of the mapped bird eggs; using a memory operator to update the memory pool; using a selection operator toselect; renewing new nests to update the quantum bird eggs and calculating the fitness of mapped bird eggs; using the memory operator to update the memory pool; using the selection operator to select; determining whether to reach a current phase iteration number; if the current phase iteration number is reached, using a purification operator to delete a similar memory element and entering into anext phase; and determining whether to reach a maximum iteration number: if the maximum iteration number is not reached, continuously carrying out iteration; otherwise, outputting all the quantum birdeggs in the memory pool, after mapping, defining as a wave reaching direction. The method possesses advantages that an operation amount is low, a convergence speed is fast and direction finding precision is high.

A fully-digital distributed scintillation crystal lifting furnace control system includes an upper computer, an intermediate-frequency heating module, a weighing module, and a crystal growth control module. A control module communicates with the weighing module to obtain a weighing signal, the control module communicates with the intermediate-frequency heating module to set a power value and read an intermediate-frequency parameter, the control module communicates with the upper computer to upload a monitoring signal and receive an operating command, and all the communication is digital and greatly improves the control accuracy of the system. The system provided by the invention is designed based on the ideal of a distributed control system. The control module and the upper computer are mutually independent in function, so that crystal growth control can get rid of software of the upper computer and normally proceed, and hidden danger brought by faults of the upper computer is prevented. Meanwhile, one upper computer can be connected with multiple lower computer control modules, so that crystal growth can be monitored in a more concentrated manner, and equipment and manpower costs are reduced greatly.

The invention discloses a fast optimization depth hash image coding method and a target image retrieval method. The fast optimization depth hash image coding method comprises the steps: building a hash image coding model for a large-scale image data set based on a greedy strategy, and generating binary codes of all images through a depth hash coding network obtained after optimization; and when target image retrieval is carried out, rapidly obtaining similar images of the same kind of a query image by calculating the Hamming distance between the query image code and the database image code. According to the fast optimization depth hash image coding method, the problems of gradient disappearance and quantization errors are better solved in combination with the neural network, and the codingperformance is better; the training process of the deep network is completed with fewer iterations, and the training speed is higher; the fast optimization depth hash image coding method can be applied to various problems with discrete constraints, and the application range is wider; and the optimization speed of the deep neural network and the retrieval performance of the generated image code are further improved, and the retrieval precision of the large image database is effectively improved.

The invention provides a method and a device for wireless communication. The method comprises the following steps: transmitting a first wireless signal by UE (User Equipment) in a first time window; and transmitting a second wireless signal in a second time window, wherein the first wireless signal contains first information, and the second wireless signal contains second information; the first information and the second information are used for determining M merged vectors; the first information is used for determining M vector groups; the second information is used for determining M coefficient groups; the M vector groups correspond to the M coefficient groups one by one; and the merged vectors are generated from the vector groups and the corresponding coefficient groups. According to the method disclosed by the invention, CSI (Customer Satisfaction Index) feedback overland is reduced on premise of ensuring the CSI feedback precision.

The invention discloses a full-connection bootstrap digital image filtering method and system. The method at least comprises the steps of extracting the minimum spanning tree of a bootstrap image; determining the spatial similarity weight on the minimum spanning tree; according to the spatial similarity weight, calculating a linear representing coefficient on an image domain; based on the linear representing coefficient and the spatial similarity weight, conducting the weighted average treatment on pixels one by one. According to the method, based on the four-adjacency relation among the pixels of a bootstrap image, the formed bootstrap image is specifically represented in such a manner that the pixels of the bootstrap image are regarded as nodes. Based on the four-adjacency relation among the pixels, a connection relationship among edges between the nodes can be established. After that, the spatial similarity weight between two adjacent pixels is measured and the weight value is assigned to an edge formed through connecting the corresponding nodes of the above two pixels. Therefore, the graph representation of the bootstrap image is formed. Based on the method and the system, the technical effect of improving the edge keep smoothing effect is at least partially realized.

The invention discloses a vector quantization method based on the normal distribution law. In the codebook generation, the feature center of a training sample set is first used as an initial codeword, and then the initial codeword is divided. When the codeword is not divided for the first time, the width of each codeword and the number of training samples belonging to each codeword are multiplied. The L codewords with the largest product are selected as the codewords to be divided, and a new codeword mean value is obtained to realize the codeword division. Every time after the codeword division, a data partitioning stage starts. In the data partitioning, the codewords are described as normal distribution, the degree of membership of each training sample to each codeword is calculated, and the cell is divided to update the codeword. Whether the quantitative distortion is convergent is judged, and if not, the data partitioning continues. If the quantitative distortion is convergent, whether the total number of the codewords has reached a certain value is judged, if not, the codeword division continues, and if so, a final codebook is output. The method of the invention can improve the accuracy of codeword division and reduce the error of vector quantization.

In a communication system, a precoder is coupled between first and second portions of each of a plurality of transmitters. The precoder is configured to receive an adaptive index representation of a scaled signal associated with the first portion of a first one of the transmitters, and to process the adaptive index representation using an adjusted precoder coefficient to generate a precompensation signal. The precompensation signal is supplied from the precoder to the second portion of a second one of the transmitters for use in precompensation of at least one signal to be transmitted by the second transmitter.

Provided is a method for implementing a high-order disturbance observer in a fixed-point DSP. The method comprises implementing the high-order disturbance observer with a cascade connection manner so as to effectively prevent influences caused by quantization errors and truncation errors in a calculating process and guarantee the operational precision of the high-order disturbance observer in the fixed-point DSP; and implementing the high-order disturbance observer in the fixed-point DSP with a manner integrating a integer-type format, a floating point formant, and an IQ format so as not to influence the operational precision of digital control quantity and to shorten the running time of a program in order to decrease the time delay of the disturbance observer and increase response speed.

The invention relates to a method for optimizing cognitive radio system parameters in a membrane structure. According to the method, by means of a mixture quantum goose swarm method in the membrane structure, the minimized transmitting power, the minimized bit error rate and the maximized data rate of a cognitive radio system are optimal simultaneously. The method includes the steps that the membrane structure is determined, the quantum position and the speed are generated, the system parameters and the mixing positions correspond to each other in a one-to-one mode, the speed and the quantum position are updated, the system parameters are formed through mapping, a fitness value is calculated, an optimal local mixing position and an optimal global mixing position of each goose are updated, the optimal global mixing positions are updated, and the optimal global mixing positions traveled by all geese of the whole quantum goose swarm are mapped to the system parameters and output from a surface membrane. The optimization problem of the system parameters with mixed disperses and continuous parameters is solved, the mixture quantum goose swarm method in the membrane structure is designed to serve as a solving strategy, and the designed method has the advantages of being high in convergence precision and convergence speed.

The invention discloses a system and a method for matched filtering of a spread spectrum signal based on frequency domain processing. In the invention, the spread spectrum signal is sent to a data caching module through an odd-even sequence gating module, data output by the data caching module is sent to a fast Fourier transform module to figure out a frequency spectrum and the frequency spectrum is sent to an odd-even sequence spectrum separation module, and a frequency spectrum of an input digital signal is obtained by an output result through a spectrum synthesis module. The frequency spectrum of the input digital signal is input into a coefficient matching module, and a digital spread spectrum signal autocorrelation result is obtained through multiplication, addition and shift operations with matching coefficients. According to the system and the method, frequency mixing and filtering are realized in a time domain, a quantization error introduced by frequency mixing and filtering during time domain multiplication can be avoided and hardware resources occupied by processing of a high-order filter are reduced.

The invention relates to an image quantization method and a storage medium. The method comprises the steps of performing region-of-interest segmentation processing on an acquired reference original image and an acquired floating original image, and determining a reference segmentation image corresponding to the reference original image and a floating segmentation image corresponding to the floating original image; determining a conversion relationship between the reference original image and the floating original image according to the reference original image and the floating original image;converting the floating segmented image by using the conversion relationship to obtain a converted segmented image; performing quantization on the reference segmentation image and the conversion segmentation image to obtain a first quantization result corresponding to the reference segmentation image and a second quantization result corresponding to the conversion segmentation image; and comparingthe first quantization result with the second quantization result, and displaying a comparison result. By adopting the method, the accuracy of an image quantization result can be improved.

The invention relates to an implantation detection circuit, an implantation detection device and an implantation detection method for a cardiac pacing system. The implantation detection circuit comprises a detection module and a voltage comparison module which are connected with a pacing circuit, the detection module is used for successively detecting the voltage of a pacing pulse before and afterelectric discharge and outputting the voltage difference signal of the pacing pulse before and after the electric discharge, the voltage comparison module is used for judging whether a voltage difference exceeds preset threshold voltage, and the voltage comparison module is further connected with the control module of the cardiac pacing system so as to transmit a comparative result to the controlmodule. Compared with traditional implantation detection technology, the implantation detection circuit of a cardiac pacemaker system of the invention does not need to accurately measure current value, electrode impedance or voltage difference value, and is conducive to lowering design difficulty and simplifying a circuit structure. Further, the detection module can be realized by detection capacitors, which have a certain connected relation, and four switches, and is high in detection accuracy, simple in structure and relatively low in power consumption.