63 results about "Variable precision" patented technology

The invention discloses a method and an apparatus for fixed-pointing the layer-wise variable precision in a convolutional neural network. The method comprises the following steps: estimating fixed-pointing configuration input to various layers in the convolutional neural network model respectively in accordance with input network parameters and a value range of input data; based on the acquired fixed-point configuration estimation and the optimal error function, determining the best fixed-point configuration points of the input data and network parameters of various layers and outputting the best fixed-point configuration points; inputting respectively the input data which is subject to fixed-pointing and an input data of an original floating-point number as a first layer in the convolutional neural network and computing the optimal fixed-point configuration point of the output data of the layer, and inputting the output result and an output result of the original first layer floating-point number as a second layer. The rest of the steps can be done in the aforementioned manner until the last layer completes the whole fixed-pointing. The method of the invention guarantees the minimum precision loss of each layer subject to fixed-pointing of the convolutional neural network, can explicitly lower space required by storing network data, and can increase transmitting velocity of network parameters.

The invention relates to a cyanobacteria biomass spatial-temporal change monitoring and visualization method based on a remote sensing image. The method comprises the following steps: (1) pre-processing the remote sensing image of a research region, and constructing a normalized difference cyanobacteria bloom index (NDI-CB); (2) optimizing characteristics of the remote sensing image by using a characteristic optimization model based on VPRS (Variable Precision Rough Set)-GID (Grey Incidence Decision), and obtaining an optimized multi-characteristic space; (3) establishing a double-weighted SVM (Support Vector Machine) classification model based on a wavelet kernel according to the multi-characteristic space, performing extraction identification and change detection on the spatial distribution information of cyanobacterial bloom, and performing comprehensive verification and precision analysis by combining field observation data; and (4) performing overlapping display on the processed remote sensing image, GIS (Geographic Information System) vector data and the field observation data, thereby realizing the analog simulation of spatial-temporal change processes and rules of erupting the cyanobacterial bloom. Compared with the prior art, the cyanobacteria biomass spatial-temporal change monitoring and visualization method based on the remote sensing image has advantages of high cyanobacteria identifying precision and reliability, and the like, and is beneficial to analyzing and judging of causes and distribution changes of the cyanobacterial bloom.

A method and apparatus for performing motion estimation in a digital video system is disclosed. Specifically, the present invention discloses a system that quickly calculates estimated motion vectors in a very efficient manner. In one embodiment, a first multiplicand is determined by multiplying a first display time difference between a first video picture and a second video picture by a power of two scale value. This step scales up a numerator for a ratio. Next, the system determines a scaled ratio by dividing that scaled numerator by a second first display time difference between said second video picture and a third video picture. The scaled ratio is then stored calculating motion vector estimations. By storing the scaled ratio, all the estimated motion vectors can be calculated quickly with good precision since the scaled ratio saves significant bits and reducing the scale is performed by simple shifts.

The invention belongs to the field of a prediction method of the urban water supply network tube explosion, and provides a novel urban water supply network tube explosion prediction method to solve the problem that the prediction precision is low and establish a tube explosion prediction model so as to provide novel study basis for urban water supply network tube explosion. Therefore, according to the technical scheme, the urban water supply network tube explosion prediction method comprises the following steps of: (1) acquiring the tube explosion rate of a tested water supply network and historical data of other relevant affect factors which affect the tube explosion rate, and taking the tube explosion rate and the historical data as the detected data; (2) establishing a water supply network tube explosion prediction model of a self-adaptive neural network heterozygosis variable precision rough set; (3) carrying out training and test analysis on the water supply network tube explosion prediction model; and (4) predicting the tube explosion rate of the water supply network by using the water supply network tube explosion prediction model which passes the test. The method is mainly applied to prediction on the urban water supply network tube explosion.

Systems and methods for processing data using an array of data processing elements that are coupled together with a variable precision interconnect. One embodiment comprises data processing elements coupled by variable precision interconnects to form a row-column array. The interconnects and / or data processing elements may be synchronous or asynchronous. The data processing elements may operate in a fixed manner, or they may be programmable, and selectable data processing elements in the array may be bypassed. The interconnects and data processing elements may be configured to handle data in a digit-serial manner, with tags for each digit identifying whether the digit is the first and / or last digit in a data word. The data processing elements may be coupled to a system bus that enables communication of data between the data processing elements and external devices and allows control information to be communicated to and from the data processing elements.

The invention discloses a similar variable precision rough set model-based knowledge pushing rule extraction method and belongs to the field of knowledge engineering. The method comprises the steps of extracting and processing user behavior data, establishing a decision table comprising condition attributes and decision attributes, obtaining the importance of the condition attributes relative to the decision attributes by utilizing an information entropy theory, and based on this, performing reduction on the decision table by utilizing the importance of the condition attributes relative to the decision attributes to obtain a reduced decision table; extracting a decision rule containing a certainty factor based on the reduced decision table; and performing verification assessment on a pushing rule, and after the rule assessment is passed, performing knowledge pushing by utilizing the rule, so that the knowledge pushing precision is improved. According to the method, the problem that the rough set model is excessively rigorous can be solved; the fault-tolerant capability of the rough set model can be improved; the method is suitable for a knowledge pushing rule extraction situation; and in addition, the high-quality knowledge pushing rule can be obtained, the knowledge pushing precision can be improved, the knowledge obtaining cost can be reduced, and the knowledge obtaining efficiency can be improved.

The invention discloses a genetic algorithm and variable precision rough set-based PET / CT high-dimensional feature level selection method. According to the method, on one hand, a chromosome coding value, a minimum reduction number of attributes, attribute dependency and the like are comprehensively considered to construct a universal fitness function framework, and different fitness functions arerealized by adjusting weight coefficients of factors; and on the other hand, for the limitation of a Pawlak rough set model, a classification error rate beta is introduced for broadening strict inclusion of lower approximation in the Pawlak rough set model to partial inclusion, so that the concept of an approximation space is perfected, the noise processing capability is enhanced, and the beta range is continuously changed to realize different fitness functions. Experimental results show that different weight coefficients greatly influence the results under the condition of consistent classification error rate; and likewise, under the condition of consistent weight coefficient, the classification error rate is increased constantly, the experimental results have relatively large difference,and a parameter combination most suitable for the method can be found according to data in the method.

The invention relates to a driving safety monitoring device and method based on car networking BSM (Basic Safety Message) information fusion and belongs to the technical field of intelligent vehicles.According to the method, on the basis of safety pre-warning of a traditional vehicle and a special vehicle, a driving safety monitoring device and method based on car networking BSM information fusion are provided, a variable precision rough set and an information entropy measuring algorithm are combined, the collision risk degree in the driving process of a general vehicle or a special vehicle is judged, and real-time pre-warning is carried out. According to the device and the method in the invention, vehicle safety operation influence factors of multiple aspects of people, vehicles and roads are fully considered, thereby improving the accuracy of collision risk identification in a complex road traffic environment. The device and the method in the invention are high in calculation speed,can meet a functional requirement for real-time pre-warning, and can be widely applied to anti-collision pre-warning and safe driving auxiliary systems and driving simulation platforms of various vehicles (including special vehicles) for evaluation and prediction.

Systems and methods for processing variable precision data using tags to identify the positions of digits within data words. One embodiment comprises a processor having internal structures that are configured to represent a variable precision data word as a variable number of digits, where each digit includes a digit value and associated tags indicative of the digit's position within the data word. The digit value may comprise an 8-bit value, and the tags may include single bits indicating whether the digit is the first and / or last digit in the variable precision word. The processor may be coupled to other variable precision devices by variable precision communication channels. The processor may be coupled to external devices that represent with fixed precision, and may use aliases to provide mappings between the variable precision data and fixed precision data, automatically adding or removing the tags associated with the digits, as necessary.

The invention discloses a suction force controllable unsaturated soil static triaxial apparatus based on internal body variable precision measurement. The static triaxial apparatus comprises a main machine part, a water circulating and pressure-stabilizing thrusting water source part and a data acquisition and processing part, wherein the water circulating and pressure-stabilizing thrusting water source part is mainly composed of two independent waterway parts formed by two thrusting pressure-stabilizing water sources, and the two independent waterway parts comprise a confining pressure control waterway for providing stable pressure for a confining pressure chamber and a monitoring waterway for precision measurement of internal body variables and suction force. The static triaxial apparatus still adopts the axis-translation technology without adopting a ceramic plate used by the traditional unsaturated instrument necessarily, and can realize precision measurement of the unsaturated soil body variables and control of the suction force in a testing process, therefore, a process that the ceramic plate is saturated is removed, the lagging problem of water drainage of the ceramic plate is solved, and the suction force in the unsaturated soil can be accurately controlled; according to the suction controllable unsaturated soil static triaxial apparatus, automatic control and measurement can be realized in the whole process, the requirements for reasonable experiment principle, high measurement precision and technical feasibility can be met, and the testing efficiency can be greatly improved.

A precision-variable laser-electron method and equipment for measuring 2D tilt angle feature that a reflector is arranged above the liquid surface or an object floating on the liquid surface, and the multiple optical reflections are used to amplify the change of tilt angle, so realizing high-precision and precision-variable measurement of tilt angle. The number of reflections and amplification coefficient can be regulated by its equipment. Its advantages are simple structure, small size, low cost and high stability and reliability.

A secure location sharing system is disclosed that allows publishers to share their locations with subscribers. Each publishing device generates, encrypts and publishes its location information with a specified precision to a distribution service for distribution to one or more subscribing devices. The distribution service provides secure distribution of the location information to the distribution list but cannot access the publisher's precise location information, which remains encrypted until decrypted at the subscribing devices. Publishers can share their location information with varying levels of precision for different subscribers. Publishers can select at least two dimensions of location information precision: spatial and temporal. For less precise location sharing, the publishing device can randomize or obscure the precise location of the publishing device. The publisher can move a subscriber or a group of subscribers to a lower or higher level of location precision by generating new key pairs at the publishing device.

The invention discloses a variable-precision optical fiber sensor, which includes a probe, a light source and an optical receiver, the light source and the optical receiver are separately connected with probe through luminiferous medium in luminiferous connection, which is characterized in that the probe is a double optical fiber probe including sensor end, light incidence end and light emergence end, lots of the double optical fiber probes are arranged in a row on the surface, every double optical fiber probe sensor end is arranged according to height difference delta H; the luminiferous medium is cable, which is composed of optical fiber bundle comprising optical fiber induced from light incidence end or light emergence end of every double optical fiber probe; the double optical fiber probe with height difference delta H is arranged in space to form a multi-row with height difference delta H'. The invention can be used for testing liquid level in inflammable and explosive fluid, which has the characteristics of simple structure, low producing cost, strong capability of anti-electromagnetism and anti-interference, and anti-corrosion.

The invention discloses a parameter-variable precision plasma antenna radiation performance simulation method and aims to realize detailed analysis on relation between plasma antenna radiation performance and plasma physical property to improve simulation precision. The method includes: firstly, determining a physical model of an antenna and incident electromagnetic wave frequency of the antenna according to a system and an operating frequency of the antenna, substituting into a Drude dispersion equation, and taking the fact that electrical conductivity can be represented by plasma relative dielectric constant values as limits to obtain plasma frequency omega pe and plasma collision frequency v by calculation; adopting CST software for setting up an antenna model which is a dispersion model, setting parameters of the dispersion model according to the plasma frequency omega pe and the plasma collision frequency v, and starting simulation to obtain an antenna radiation performance result; finally checking the antenna radiation performance result to judge whether the result meets requirements or not, and if not, re-determining a group of parameters and repeating the steps.

The invention provides a fatigue condition monitoring method and device, wherein the method comprises the following steps of building a similar decision making rule used for judging the fatigue condition according to brain wave feature data by using the brain wave feature data as condition attributes in advance, using fatigue judging parameters as decision making attributes and using a variable-precision rough set model; then, collecting the brain waves of a user in real time; next, calculating the attribute values of the brain wave feature data corresponding to the brain waves; finally, determining the fatigue condition of the user according to the attribute value of the brain wave feature data and the similar decision making rule; and completing the real-time monitoring on the fatigue condition of the user. The method and the device provided by the invention have the advantages that the method is simple; the realization is easy; the cost is low; and meanwhile, the accuracy and the real-time performance are high.

Systems and methods for processing data using an array of data processing elements that are coupled together with a variable precision interconnect. One embodiment comprises data processing elements coupled by variable precision interconnects to form a row-column array. The interconnects and / or data processing elements may be synchronous or asynchronous. The data processing elements may operate in a fixed manner, or they may be programmable, and selectable data processing elements in the array may be bypassed. The interconnects and data processing elements may be configured to handle data in a digit-serial manner, with tags for each digit identifying whether the digit is the first and / or last digit in a data word. The data processing elements may be coupled to a system bus that enables communication of data between the data processing elements and external devices and allows control information to be communicated to and from the data processing elements.

The invention discloses a variable-granularity feature sampling method based on a multimedia sensor, and belongs to the field of energy consumption and storage of multimedia sensors. In the sampling process of the multimedia sensor, the residual energy of the sensor is estimated by the period, an energy-saving mode is switched by a membership threshold function, and a high-precision mode is switched by a high-precision measurement threshold function to realize the sampling mode of variable-precision sampling, thus changing the active and dormant time in the period, achieving energy-saving andhigh-precision sampling goals, and enhancing the working time and value of the sensor. Through adoption of the method, the sensing accuracy and the network lifetime can be balanced effectively; a balance is achieved between the two; and meanwhile the measurement accuracy and the reduction of energy consumption are ensured.

Systems and methods for processing data using an array (11) of data processing elements (13) that are coupled together with a variable precision interconnect (14). One embodiment comprises data processing elements coupled by variable precision interconnects to form a row-column array (11). The interconnects (14) and / or data processing elements (13) may be synchronous or asynchronous. The data processing elements (13) may operate in a fixed manner, or they may be programmable, and selectable data processing elements (13) in the array may be bypassed. The interconnects (14) and data processing elements (13) may be configured to handle data in a digit-serial manner, with tags (25, 26) for each digit identifying whether the digit is the first and / or last digit in a data word. The data processing elements may be coupled to a system bus (54, 55) that enables communication of data between the data processing elements and external devices and allows control information to be communicated to and from the data processing elements.

The invention discloses variable-precision screw-extruding additive manufacturing equipment with stirring and anti-blocking functions. The equipment comprises a three-dimensional movable platform andextrusion units arranged on the three-dimensional movable platform. Each extrusion unit comprises a first motor, a gearbox, a storage bin, a stirring part, a screw, a plunger, a cylinder and a discharging nozzle, wherein the first motor is connected with the gearbox; the bottom of the gearbox is connected with one end of the storage bin; the stirring part is arranged in the storage bin; the screwcomprises a fixing part and a threaded part and penetrates through the storage bin and the cylinder, one end of the screw is connected with the stirring part through the fixing part, the other end ofthe screw is connected with the plunger through the threaded part, an output shaft of the gearbox is connected with the fixing part, the bottom of the storage bin is connected with one end of the cylinder, and the bottom of the cylinder is connected with the discharging nozzle. With the adoption of the equipment, the problems of low efficiency, high material cost and the like in traditional additive manufacturing are solved, and various materials such as engineering plastics, building materials, high-viscosity energetic materials, metal slurry, and ceramic slurry can be formed.

The invention provides a data compression device for a neural network processor, a supporting data decompression device, and a neural network processor with variable precision. The data compression apparatus includes a rounding unit, a first comparator, a second comparator, and a first multiplexer for converting m bits of data to be compressed into n bits of compression results O, n (m. Through the data compression device, the neural network processor can adopt different data precision to participate in calculation at different layers of the neural network, and on the premise of ensuring the calculation precision, the on-chip storage amount is reduced, the data transmission energy loss is reduced, and the device is particularly suitable for the applications with limited on-chip resources.

The invention relates to a full-drive-by-wire four-wheel steering angle measuring system suitable for operating lever control. An angle sensor adopted by the steering angle measuring system is a non-contact sensor free of wire connection. Angle detection is conducted through optical media, and measurement is free of influences of temperature and magnetic fields. The steering angle measuring system is simple in structure and easy to integrate, effective match with an operating level of full-drive-by-wire four-wheel steering of operating level control and wheel assemblies can be achieved, the operating lever is provided with one steering angle measuring system in a matched mode, and the four wheel assemblies are provided with four steering angle measuring system in a matched mode. The steering angle measuring system is high in measurement precision, the signal transmission delay time is short, initial position correction and electronic limiting can be conducted on angles of wheels, the vehicle safety is improved through the multiple steering fault-tolerant capability, and the manufacturing cost is reduced through variable-range and variable-precision measurement.

The invention provides a rough set expansion personalized recommendation method based on the variable precision tolerance relation. By means of two object indistinguishability characteristic analysis which is more consistent with the statistical significance, a core set is acquired by means of the discernable matrix, and discernable matrix elements containing core attributes are emptied. A reduction set is initialized into the core set, corresponded reduction and union are performed to attributes appearing mostly of discernable matrix non-empty element with those of the reduction set, fresh reduction is acquired, the elements with the attributes in the discernable matrix are emptied, and extraction of the reduction set is implemented. Focusing on each reduction, a knowledge tree is established, pruning conditions are determined, special pruning can be performed according to the pruning conditions, calculation complexity is reduced, effective rules are acquired, a rule set is formed, and data excavation is implemented effectively. According to the condition attribute value of the result to be recommended, rule match is performed in the rule set, and personalized recommendation can be implemented.

A digital image automatic labeling method based on uncertainty analysis, including the steps of extracting image features based on a deep convolutional neural network, constructing an image automaticlabeling system based on a variable precision neighborhood rough set, and labeling unlabeled images. The method includes the following steps: collecting the image data and labeling to obtain a training set, and extracting a feature vector of the image through the deep convolutional neural network; obtaining a classification model based on the neighborhood estimation class conditional probability density; in prediction, extracting image features, and estimating the position of the image to be classified by using upper and lower approximation concepts of the rough set; directly judging the membership of the labels for the images located in positive and negative domains, and judging the images in the boundary domain by using a Bayesian decision rule. According to the digital image automatic labeling method based on uncertainty analysis, the position of images to be labeled in the sample space are estimated by introducing upper and lower approximation concepts of the rough set, the error prediction rate of the irrelevant labels is reduced, and the problem of uncertainty existing between the underlying image feature and the high level semantics in image automatic labeling is solved.

The invention provides a variable precision adjustment method for a playing content of a cinema. The method comprises the following steps of inputting a playing action precision value F; obtaining a progress value P of the current playing content of the cinema, wherein the progress value P comprises an integer part of a summing value for converting the total time of the current playing progress into seconds and a decimal part MS of the summing value for converting the total time of the current playing progress into seconds; computing to obtain a current action serial number ID according to the playing action precision value F and the progress value P of the current cinema playing content, and realizing adjustment of the control precision for a dynamic cinema according to the current action serial number ID, wherein a computational formula of the current action serial number ID is ID=S*F+MSID. According to the method, different precisions are required according to the playing content, adjustment of the precision F value is needed only, so that the table can complete F groups of action in one second, and the adjustment of the control precision for the dynamic cinema is realized.

The invention discloses a rangeable seismic ground penetrating radar trolley with a canopy, which comprises a vehicle body, a console, a radar box, a radar lifting control structure, a rangefinder system, an anti-seismic wheel structure frame, an anti-seismic wheel and a battery. The ground penetrating radar (GPR) can move freely under various road conditions by the vehicle through the trolley, and the GPR can work steadily. The vibration energy can be effectively filtered out to ensure the stable operation of the trolley in the face of some uneven concave soil or debris on the surface, whichmay cause turbulence when vibration is absorbed in the form of anti-vibration wheel set. The trolley has a ranging function, utilizes the torsion generated by the contact between the intermediate shock absorbing wheel and the ground, generates the rotating turns through the meshing of the gears and records the distance of the trolley movement; In addition, the real-time fitting between the variable precision gear of the rangefinder and the gear of the axle of the rangefinder is used to ensure the stability of the distance record of the trolley motion.

A variable-precision distributed arithmetic (VPDA) multi-input multi-output (MIMO) equalizer is presented to reduce the size and dynamic power of 112 Gbps dual-polarization quadrature phase-shift-keying (DP-QPSK) coherent optical communication receivers. The VPDA MIMO equalizer compensates for channel dispersion as well as various non-idealities of a time-interleaved successive approximation register (SAR) based analog-to-digital converter (ADC) simultaneously by using a least mean square (LMS) algorithm. As a result, area-hungry analog domain calibration circuits are not required. In addition, the VPDA MIMO equalizer achieves 45% dynamic power reduction over conventional finite impulse response (FIR) equalizers by utilizing the minimum required resolution for the equalization of each dispersed symbol.

A processor selectively adjusts the precision of data for different functional units. Specified functional units of the processor, such as shader processing unit of a graphics processing unit (GPU) include a zeroing module to store, based on the states of corresponding precision flags, a data value of zero at specified portion of an input and / or output data operand. The functional unit then processes the data including the zeroed portion. Because a portion of the data has been zeroed, the functional unit consumes less power during data processing. Furthermore, the precision flags are set such that the reduced precision of the data does not significantly impact a user experience.

A method of encoding variable-precision numeric types that includes determining a fixed numeric storage size used by a computer system, encoding numeric values that do not fit into the fixed size in a variable-precision numeric value allocated on a program heap and generating a reference to point to the variable-precision numeric value on the program heap.

Disclosed is a GBuffer data encoding and decoding method suitable for deferred shading. At the time of encoding, data to be encoded are first converted into linear spatial data, then the linear spatial data are split into fixed data and optional data, and the optional data part in the linear spatial data is stored using an ALpha channel. Correspondingly, at the time of decoding, the split optionaldata are added to originally corresponding data bits. According to the invention, the data bits on the ALpha channel is used to increase the storage of more bits of data, so as to improve the accuracy of the encoded data. When high-precision encoding is not required, encoding can be done directly in the original manner, and no additional data storage amount will be increased because of precisionrequirements. The method provided by the invention uses only 22 bits in the process of octahedral normal encoding and has an effect exceeding the effect of the existing 24-bit BestFitNormal encoding.The method of the invention is particularly suitable for encoding and rendering of small bandwidth video cards.