41 results about "Float Value" patented technology

The invention is an ECN-based forecasting and verifying congestion control method, firstly using the difference between two adjacent packets to calculate forecast parameters of a current packet, by this recurring to the corresponding forecast queue length of, and able to refer to the congestion conditions of nodes to correct the forecast length as a common forecast threshold value for all the packets; taking 2 x the first packet delay as signal transmitting time and all the packets as forecast packets in order to find a forecast value of average queue length of all the packets, a receiving end compares the difference between the forecast value and the forecast threshold value with predicted floating value of the forecast value, so as to be able to send a confirm packet increasing, decreasing or keeping the length of congestion window to the source end, convenient for the source end to regulate the packet transmitting speed. This receiving end driving method has the advantage of reducing loss rate and average packet delay at the same time, has stale and higher network utilization ratio and can also forecast the network's congestion conditions.

The invention discloses a transaction index abnormality monitoring method based on a deep learning model LSTM. The method comprises the steps that 1 transaction index historical data are collected and processed; 2 the LSTM model with transaction index time series pre-determining is trained for the historical data processed in the step 1; 3 through the LSTM model of the step 2, the normal index interval floating value is calculated; and 4 the current index value is predicted through the LSTM model of the step 2, and abnormality capturing is carried out on the actual index value according to the normal index interval floating value of the step 3. According to the invention, the time series index is predicted through the deep learning model LSTM; the predicted current index value is accurate; the pre-determining accuracy of irregular time series indexes is greatly improved; subsequent abnormality capture is accurate; the output result predicted by the LSTM model is combined, and a residual fitting formula and a logistic regression model algorithm are used to calculate the normal interval floating value; abnormality capture is accurate; and the efficiency is improved.

The invention discloses a three-dimensional reconstruction method based on deep learning, and the method comprises the steps: (1), reconstructing a complete three-dimensional shape of a target througha constrained potential vector of an input image, carrying out the mapping between a learning part and the complete three-dimensional shape, and achieving the three-dimensional reconstruction of a single depth image; (2) learning intermediate feature representation between the three-dimensional real object and the reconstruction object so as to obtain a target potential variable in the step (1);and (3) converting the predicted voxel floating value in the step (1) into a binary value by using an extreme learning machine to complete high-precision reconstruction.

The invention discloses a self-attention multi-core maximum mean difference-based transfer learning speech enhancement method, which comprises the following steps: extracting GFCC features from original speech and using the GFCC features as input features of a deep neural network; calculating ideal floating value masking of Fourier transform domain by using noisy speech and clean speech information, and using the ideal floating value masking as the training target of the deep neural network; building a speech enhancement model based on deep neural network; building a self-attention multi-coremaximum mean difference-based transfer learning speech enhancement model; training the self-attention multi-core maximum mean difference-based transfer learning speech enhancement model; inputting frame-level features of the noisy speech in the target domain, and rebuilding enhanced speech waveform. By adding a self-attention algorithm to the front end of the multi-core maximum mean difference andby minimizing the multi-core maximum mean difference between features noticed by the source field and features noticed by the target domain, transfer learning of the unlabeled target domain is realized, and the speech enhancement performance is improved. The method of the invention has a good application prospect.

A circuit for generating a random output value is disclosed that comprises: a bistable circuit having two stable states in which a 0 or a 1 is output and having a balanced metastable state in which a floating value between 0 and 1 is output, said bistable circuit resolving from said metastable state to one of said stable states on being switched on, said state depending on a voltage level at a port on said bistable circuit; a voltage level control circuit for controlling a voltage level at said port on said bistable circuit; a time measuring circuit for measuring a switching time taken for said bistable circuit to switch from said metastable state to one of said stable states following switch on; and control logic for controlling said time measuring circuit, said voltage level control circuit and a switching off and on of said bistable circuit, said control logic being adapted to perform a following sequence: control said voltage level control circuit to set a predetermined voltage level at said port on said bistable circuit, switch said bistable circuit on, detect a measured switching time, and turn said bistable circuit off and if said measured switching time is longer than a predetermined value, output said resolved stable state of said bistable circuit as said random output value.

The invention provides an Internet of Water based remote monitoring method, system and client for elderly people living alone. The method includes acquiring current water consumption volume; acquiring the time segment of the current time point; metering the detection water consumption volume from the start time point to the current time point of the time segment and calculating the difference value between a second time segment water consumption volume threshold and a threshold floating value; detecting whether the water consumption volume is greater than the sum of a first time segment water consumption threshold and the threshold floating value, sending prompt information to the connected remote client if water consumption volume is greater than the sum of the first time segment water consumption threshold and the threshold floating value the or judging whether the current time point is in the next time segment or not if the water consumption volume is not greater than the sum of a first time segment water consumption threshold and the threshold floating value; metering the water consumption total volume of the last time segment if the current time point is in the next time segment and judging whether the water consumption total volume is smaller than the difference value of the water consumption volume threshold of the second time segment corresponding to the last time segment and the threshold floating value; sending the prompt information to the client if the water consumption total volume is smaller than the difference value of the water consumption volume threshold of the second time segment corresponding to the last time segment and the threshold floating value or updating the time segment water consumption volume threshold of the corresponding time segment according to the water consumption total volume of the last time segment of the last day and the water consumption total volume of the last time segment of the day. Therefore, low-cost and high-accuracy remote monitoring of life state of the elderly people is realized and safety condition of the elderly people can be controlled.

The present invention relates to a guide float valve plate which comprises a guide float valve which is a symmetrical quadrangular guide float valve, its valve cover is a quadrangular from with symmetrical diagonal liens, the included angle of long edges is directed against overflow weir, under the valve cover more than three valve legs are mouthed, front valve leg length is longer than rear valve leg length, so that when the float value is completely opened, between valve cover and valve hole plate an included angle toward overflow weir is opened, said included angle is 2-10 deg, and on the valve cover can be mounted the guide hole and guide plate. Said invention can effectively reduce liquid level gradient on the plate, raise plate flux, raise mass transfer efficiency and reduce resistance of plate.

The invention relates to a transmission controlling method and a device for digital television transport stream, which make random floating to PID value of an appointed transport stream when an uplink port of a digital television system reuses a program and adopts the floated value as the actual PID value of transport stream; the uplink port transports the floating value of PID to a receiving equipment by an encryption mode; the receiving equipment analyzes random rules and obtains the actual PID value of the transport stream according to the appointed PID of the transport stream. The illegal receiving equipment cannot obtain the floating random number of the PID value and receive the program without knowing the encryption mode, a cryptographic key and transmission rules.

This invention relates to electricity power system jump judging method, which adopts the following bases: A, protruding variables starting; b, )P-0.2Sí¦PS1 means front power larger than set value PSI; c, PtíœPS2, fault power smaller than set value PS2; d, three-phase current IíœIS1, current smaller than operation current; e, current variable satisfies the base larger than current float gate; f, tí¦tS1, determining the above data time lag, wherein the PSI smaller than transmission power value; PS2 is larger than zero and the maximum floating value with operation current IS1 larger than open circuit current; the UN is ratio voltage.

The invention discloses an intelligent recognition system for an optimal charging current and a method thereof. The intelligent recognition system comprises a trial current setting module, an optimal current selecting module and an output control module, wherein the trial current setting module is used for setting a charging current of a mobile phone, and the charging current of the mobile phone is tried by the optimal current selecting module; the optimal current selecting module is used for detecting the floating value of the current charging voltage at intervals of set time, if the floating value is greater than a set value, other charging currents are selected from the charging currents set by the trial current setting module, and if all the charging currents set by the trial current setting module are detected, the charging current of which the floating value of the charging voltage is minimum is selected as the optimal charging current; the output control module is used for charging batteries of the mobile phone according to the result selected by the optimal current selecting module. By using the intelligent recognition system and the method, the problems that the batteries can not be charged caused by insufficient charging current and the service life of the batteries is shortened by charging under the condition of overcurrent charging current can be solved effectively.

The invention relates to a method and device for quantising the floating value of a pixel of an image by rounding either to a lesser whole number, to a greater whole number, or to the whole number closest to this floating value. The method is characterised in that the selection of rounding this floating value is determined based on a test value.

A method and system for a GCF repo swap transaction includes generating an index using a limited set of GCF contracts, and using the index value as a value in a GCF repo swap. The index value may be used as a variable or floating value in the swap, or may be used at the fixed value in the swap.

The present invention provides a method for assessing reading ability grades, which at least includes the following steps: obtaining a user's grade, determining a benchmark value of ability according to the user's grade; obtaining user reading data, and determining a floating value of ability according to the user's reading data; obtaining user Access data and participation activity data, determine the reading enthusiasm value according to the user access data and participation activity data; sum the ability benchmark value, ability floating value, and reading enthusiasm value to obtain the reading ability level value. The invention also provides a reading ability grade evaluation system.

The invention discloses a method for judging the startup of an interphase current mutation amount through a floating threshold. The realization method is that the floating value is increased on the basis of the fixed value of the starting current of the sudden change amount, and the starting threshold of the sudden change amount of the current between phases is taken as the starting threshold value. When the instantaneous value of the interphase current abrupt change is greater than the threshold of the interphase current abrupt change, the interphase current abrupt change element is allowed to start. Using the method of judging the sudden change of current between phases by floating threshold, the frequent start-up of the protection device can be prevented when the error of the system frequency relative to the power frequency exceeds +/- 2 Hz during operation.

The invention discloses a method and a system for resource scheduling, which are used to solve the problems that the prior method for resource scheduling requires an operator to make sure of resource scheduling each working day, and the implementation is more troublesome. The method comprises the following steps that: the system for resource scheduling captures information on operation of a resource unit in each period, information on the operating status of the resource unit in each period, information on operation time in each run of the resource unit, information on free time after each run, and information on floating value of the free time after each run of the resource unit in each period, information on maximum total time and minimum total time from the first run to the last run of the resource unit, information on maximum free time information and minimum free time of the resource unit; the resource scheduling system schedules the resource unit according to the information.

The invention provides a comprehensive cooling testing working method for vehicle silencing. The method comprises the steps of S1, after tail gas current enters a second chamber through an air inlet pipe, acquiring a sound wave signal of the air inlet pipe by a first standing wave tube; S2, after a single-chip microcomputer receives the sound wave signal of the first standing wave tube, performinganalysis and calculation on the signal of each standing wave tube; S3, mounting a temperature sensor in a silencing device for performing real-time acquisition on an inner air current temperature; S4, sampling the temperature value of a cylinder body by the temperature sensor; S5, determining a voltage floating value through comparing an input voltage change reference signal with a feedback output detecting signal; and S6, if cylinder body temperature reduction cannot be realized because of a fault of a circulating water pump, after the temperature value which is acquired by the temperature sensor exceeds a certain threshold, transmitting a signal to a processor through an overheat protecting circuit, and performing real-time monitoring on operation of the circulating water pipe through acorresponding protecting circuit. The comprehensive cooling testing working method has advantages of ensuring stable operation of a water pump, and prolonging service life of a circulating refrigerating system.

Methods and apparatus are provided for estimating parameters, i.e. ambiguities, derived from GNSS signals. Observations of GNSS signals are obtained from each of a plurality of GNSS satellites (120). The observations are fed to a filter having a state vector at least comprising a float ambiguity for each received frequency of the GNSS signals (140). The filter estimates a float value for each float ambiguity of the state vector. Integer values are assigned to at least a subgroup of the estimated float values to define a plurality of integer ambiguity candidate sets (160). A first number of candidate sets is selected having a quality measure better than a first threshold, wherein the first threshold is determined based on a reference quality measure of a reference candidate set (180). A weighted average of the selected candidate sets is formed, each candidate set weighted in the weighted average based on its quality measure (200). Ambiguities of the weighted average can be used in subsequent operations to aid in determining a position of the receiver or can be used to prepare data, e.g., in a network processor that can be used to augment position information of a rover.

Methods and apparatus are provided for estimating parameters, i.e. ambiguities, derived from GNSS signals. Observations of GNSS signals are obtained from each of a plurality of GNSS satellites (1120). The observations are fed to a filter (1140) having a state vector at least comprising a float ambiguity for each received frequency of the GNSS signals. The filter estimates a float value for each float ambiguity of the state vector. Integer values are assigned to at least a subgroup of the estimated float values to define a plurality of integer ambiguity candidate sets (1160). A quality measure is determined for each of the candidate sets. The best quality measure of the candidate sets is determined. An expectation value (1170) of the candidate set having the best quality measure is determined. An error measure as a ratio of the best quality measure to the expectation value is determined. The quality measures of the candidate sets (1180) is adapted as a function of the error measure. A weighted average of a subgroup of the candidate sets on the basis of the adapted quality measures is formed, wherein at least one of selecting the subgroup of the candidate sets and the weighting of each candidate set in the weighted average is based on the adapted quality measure (1200). Ambiguities of the weighted average can be used in subsequent operations to aid in determining a position of the receiver or can be used to prepare data, e.g., in a network processor that can be used to augment position information of a rover.

Methods and apparatus are provided for estimating parameters, i.e. ambiguities, derived from GNSS signals. Observations of each of received frequencies of a GNSS signal from a plurality of GNSS satellites are obtained for a plurality of instances in time (3120). The time sequence of observations is fed to a filter to estimate a state vector comprising float ambiguities, wherein each float ambiguity constitutes a non integer estimate of an integer number of wavelengths for a received frequency of a GNSS signal between a receiver of the GNSS signal and the GNSS satellite from which it is received and wherein the float ambiguities of the state vector are updated over time on the basis of the observations (3140). The occurrence of an interruption in tracking of at least one signal of a satellite is determined (3121). The float ambiguity of the state vector for the at least one signal for which an interruption in tracking occurred is maintained at the value before the interruption in tracking occurred (3122). Integer values are assigned to at least a subgroup of the estimated float values to define a plurality of integer ambiguity candidate sets (3160). A quality measure is determined for each of the candidate sets. A weighted average of the candidate sets is formed (3200). Ambiguities of the weighted average can be used in subsequent operations to aid in determining a position of the receiver or can be used to prepare data, e.g., in a network processor that can be used to augment position information of a rover.

The invention discloses a layered sculpturing method. The layered sculpturing method comprises the following steps: S1, acquiring a sculpturing model; S2, selecting an original point, carrying out point decomposition on the surface layer of the sculpturing model, and establishing a coordinate according to the concentrated decomposed points of surface layer points; S3, establishing a reference model in a coordinate point in the step S2, wherein the reference model consists of not less than one reference point, and each concentrated decomposition point of the surface points corresponds to one reference point of the reference model; S4, generating a feeding value according to the corresponding relationship between each decomposition point and the reference point; S5, dividing the intervals of the decomposition points according to the feeding value, wherein the maximal difference of the feeding values of every two decomposition points in each interval is not greater than the preset floating value; S6, establishing a sculpturing path aiming at each interval; and S7, sculpturing a blank body according to the sculpturing path. According to the invention, the blank body is sculptured in a layered mode, and the blank body scrappage rate is reduced.