70 results about "Exponential weighting" patented technology

The invention discloses an intensive learning based urban intersection passing method for a driverless vehicle. The method includes a step 1 of collecting vehicle continuous running state informationand position information through a photographing method, the vehicle continuous running state information and position information including speed, lateral speed and acceleration value, longitudinal speed and acceleration value, traveling track curvature value, accelerator opening degree and brake pedal pressure; a second step of obtaining characteristic motion track and the velocity quantity of actual data through clustering; a step 3 of processing original data by an exponential weighting moving average method; a step 4 of realizing the interaction passing method by utilizing an NQL algorithm. The NQL algorithm of the invention is obviously superior to a Q learning algorithm in learning ability when handling complex intersection scenes and a better training effect can be achieved in shorter training time with less training data.

A novel and useful channel tracking mechanism operative to generate channel estimate updates on blocks of samples during reception of a message. The tracking mechanism is based on the weighted recursive least squares algorithm and implements the estimation process by recursively updating channel model parameters upon the arrival of new sample data. The mechanism is operative to update channel estimate information once per sample block. An interblock exponential weighting factor is also applied. The block length is chosen short enough to enable good tracking performance while being sufficiently long enough to minimize the overhead of generating preliminary decisions and of updating precalculated tables in the equalizer. The method of the invention can be performed in either hardware or software. A computer comprising a processor, memory, etc. is operative to execute software adapted to perform the channel tracking method of the present invention.

The invention provides a pipeline leakage weighted positioning method and device based on pressure waves and sound waves. The method comprises the following steps: obtaining a signal transmission time difference obtained through multiple instruments and multiple methods; and carrying out exponential weighting processing on the signal transmission signal difference by adopting an iterative method to obtain positioning results. The device comprises an obtaining module and a calculating module. According to the pipeline leakage weighted positioning method and device, comprehensive positioning precision for multiple signal sources can be remarkably improved by an exponential weighting algorithm established based on physical characteristics of a signal, and the pipeline leakage weighted positioning method is superior to a weighted positioning method based on probability statistics in requirements on sample number, positioning inaccuracy, calculated amount and the like.

The invention provides a KF (Kalman Filter) tracking method based on fading memory exponential weighting. The method comprises the following steps: a state error covariance matrix P and a systematic process noise matrix are acquired; an estimated predictive state parameter value shown in the description of a moving object at the moment k is calculated, and innovation covariance C0,k at the momentk is calculated; innovation gamma k at the moment k is calculated, an estimated innovation covariance value shown in the description at the moment k is calculated, weighting coefficient beta k at themoment k is calculated, and the fading factor lambda k at the moment k is further calculated; a predictive state error covariance matrix Pk|k-1 and Kalman gain Kk at the moment k are calculated, and an estimated state value shown in the description and a state error covariance matrix Pk are further calculated, wherein a calculation method for the estimated innovation covariance value at the momentk is shown in the description, and the weighting coefficient [beta i] decays following the law of negative exponent. The problem of poorer precision of the traditional windowing average method for calculating innovation residual vector estimation is solved, and innovation residual estimation precision is improved effectively, so that the method has higher precision and robustness.

The invention relates to a human health evaluation method based on a physiological health index, comprising the following steps: S1, measuring the human body to be evaluated, and obtaining actual measured values ​​of selected human body indexes; S2, normalizing the actual measured values ​​of each index processing to obtain the normalized value of each index; S3, assigning weights to each index; S4, determining the age and gender weight functions; S5, obtaining the LW index through linear weighting, and then obtaining the TOPSIS index by adjusting the TOPSIS model, and finally obtaining the index weighted Human body's physiological health LW&TOPSIS index. The present invention effectively fuses a plurality of physiological parameter index values ​​reflecting the health status of the human body, and gives the health evaluation result by the way of the human health index, which not only makes the evaluation result of the health status of the human body simpler and more intuitive, but also gives the evaluation result The results can enable non-professionals to clearly and clearly understand their own health status, and have important guiding significance for the adjustment of human body-related life activities and even the prevention and treatment of diseases.

The invention discloses a method for detecting the flow abnormity of a wireless sensor network based on a GM model. The method employs the GM (1, 1) model, is small in amount of used historical data, is quick in building speed of a model, is accurate in prediction value, and is very suitable for the condition that the node energy and calculation capability of the wireless sensor network are limited. enabling a historical modeling data quantity to be fixed through employing a sliding window in a proper size, thereby guaranteeing the quickness of modeling and also guaranteeing the latest effectiveness of historical data; optimizing albinism differential equation solving initial conditions of the GM (1, 1) model, and enabling the prediction value to be more accurate; generating a flow prediction value, finally used for abnormal judgment, at the next moment through the exponential weighting mean of the former L predication values, thereby introducing certain inertia to the prediction of flow. When an abnormal flow happens, a normal flow prediction model cannot be changed easily, but a normal flow prediction value can be obtained better, and the flow abnormality can be detected more easily.

The invention discloses a synchronous correction method based on an IEEE1588 clock model, and belongs to the technical field of communication. The method comprises the following steps: (1) acquiring the moment point of the IEEE1588 clock model to obtain each timestamp; (2) acquiring the relation between the time of a server and the time of a client through the IEEE1588 clock model, and acquiring the relation among the timestamps in Sync message sending and receiving processes in combination with a PTP (Picture Transfer Protocol); (3) performing difference operation based on the relation among the timestamps; and (4) acquiring the asymmetrical offset of an nth exchange process according to a first-order difference value given in the step (3), moving an average filter through exponential weighting in order that value theta is close to zero, and finishing correction between the server and the client. According to the method, a conventional thought is jumped out, and compensation is performed once again on the basis of the IEEE1588 clock model, so that the problem of end-to-end time non-synchronization is solved, and end-to-end interaction time becomes symmetrical.

A method and apparatus for polarity detection. The method includes applying a band-pass filter to an impulse response of a loudspeaker, applying an exponential weighting to the band-pass filtered impulse response, wherein the exponential decay parameter is related to the higher corner frequency of the band-pass filter, finding the maximum peak in a waveform of sampled impulse responses, and detecting the connection polarity of the maximum peak as the polarity of the peak.

The invention relates to the field of rotating equipment fault diagnosis, and discloses a state monitoring method and system based on multivariable state estimation, and the method comprises: A), collecting a working condition parameter signal and a vibration signal of equipment, and obtaining an m*n-dimensional feature parameter matrix; B) calculating the importance of each feature by adopting aGBDT method, and determining a feature selection number q according to the importance; C) calculating a state estimation value of the equipment by adopting a Mahalanobis distance method; and D) establishing an equipment exception judgment mechanism by adopting an exponential weighted average method, and judging the state condition of the equipment. According to the invention, the key parameters capable of reflecting the equipment state can be effectively selected, characteristic dimensions are reduced, state estimation time is shortened, interference of invalid characteristics is avoided, a Mahalanobis distance method is adopted to fuse multi-dimensional characteristic parameters into an estimated value, state estimation is carried out through an exponential weighted average method, fluctuation caused by random errors is reduced, instantaneous burst capacity can be absorbed, timeliness is high, and accuracy is high.

The invention discloses a modified TCPW congestion control method in a wireless network. After a specified data receiving terminal receives a data packet each time, a timestamp is stamped in a reply packet corresponding to the data packet, after a data sending terminal receives the reply packet, the arrive time of the current reply packet is not recorded, but the timestamp on the reply packet is extracted, so as to obtain the timestamp difference, namely the sending time of the last data packet on the sending route, in the next arrived reply packet, the size of the arrived data packet and the sending time of the data packet on the sending route are respectively subjected to averaging at the sending terminal, and then a bandwidth sampling value is obtained according to the specific value of average values. Finally a bandwidth estimated value is smoothly obtained through an exponential weighting moving average filter according to the obtained bandwidth sampling value. The method is more accurate to estimate a current practical available bandwidth value and effectively improves network QoS.

The invention discloses a channel distribution and user correlation strategy based on an AMAB model. The strategy includes the steps that decision probability sequences are distributed at equal probability through APs and STAs; each AP selects a channel according to the corresponding probability sequence, information such as the average data arrival rate, the time relay and the throughput capacity of the STAs is counted, and the gain of the currently-selected channel is calculated; the APs calculate accumulated price parameters; the APs calculate new probability sequences according to an index average weighting strategy; each STA selects the corresponding AP for correlation according to the corresponding probability sequence, information such as the average data arrival rate, the time relay and the throughput capacity is counted, and the gain of the currently-correlated AP is calculated; the STAs calculate an accumulated price function; the STAs calculate new probability sequences according to an index weighting average strategy and conduct calculation till the strategy convergence is optimal. By the adoption of the channel distribution and user correlation strategy based on the AMAB model, the strategy can converge at nash equilibrium, the optimal solution is acquired, and the problem of large network time delay caused by same-channel interference in an intensive scene can be solved effectively.

The invention pertains to a method of manufacturing parts produced with a manufacturing device, based on the analysis of at least one statistical indicator representative of a characteristic dimension of the parts, according to which: a) in the course of time several samples are collected, each sample comprising several parts produced with the manufacturing device; b) the characteristic dimension of each part of the sample is measured; c) for each sample collected a weighted mean and a weighted standard deviation of the characteristic dimension are calculated according to an exponential weighting on the basis of a mean and standard deviation of the characteristic dimensions measured on the parts of said sample, of weighted means and of weighted standard deviations of the characteristic dimension which are calculated for previously collected samples; d) for each sample collected a value of the statistical indicator is calculated on the basis of the weighted mean and of the weighted standard deviation thus calculated; e) a value of the statistical indicator thus calculated for the sample collected is compared with a reference value to detect a possible deviation; f) the manufacture of the parts is steered as a function of the results of the comparison by fitting the manufacturing device adjustment parameters to optimize the deviation between the value of the statistical indicator and the reference value.

Aiming at the problems that a data source is single, a sequence model difficultly captures a lane changing intention in a long sequence range and long-term dependence exists in lane changing intentionidentification, the invention provides a long-term short-term memory network vehicle lane changing intention identification model under a time information weighting index loss function. The method comprises the steps: firstly, conducting a highway driving experiment through a driving simulation cabin and an eye tracker, and collecting vehicle operation data and driver eye movement data; constructing a vehicle lane changing intention identification model in a highway environment based on an LSTM structural unit, and optimizing the model weight through a proposed index loss function based on time information weighting; and finally, verifying the proposed model by using the vehicle operation data and the driver eye movement data and comparing the proposed model with other models, wherein the lane changing identification accuracy of the proposed model is 96.78%, the precision is 95.72%, the recall rate is 95.83%, and the F1 value is 95.73%. The LSTM network has good resolution capabilityfor a long-sequence lane changing intention identification process, and the proposed loss function has a good effect on model weight optimization.

The invention discloses a ballistic trajectory formation method based on exponential weighting attenuated memory filtering. During the ballistic trajectory formation process, errors generated during the trajectory model linearization process are reduced by a state deviation Kalman filter equation, and filter divergence caused by errors of the trajectory model is inhibited by the exponential weighting attenuated memory filtering. Thus, reliable filter data is provided for a trajectory prediction system. According to the invention, filter divergence can be inhibited effectively, and system stability can be raised.

The invention discloses a system and a method for estimating noise of compass optical fiber gyroscopes for ships on line. The system for estimating on line comprises three optical fiber gyroscopes, three accelerometers, a signal collecting part, a DSP (Digital Signal Processor) navigation calculating part, a power supply part and an information displaying part, wherein the signal collecting part is used for collecting output signals of the gyroscopes and the accelerometers in real time, the collected output data of the gyroscopes and the accelerometers is filtered, interference influence on collection data caused by ship motion is removed and only noise interference items of the optical fiber gyroscopes are reserved, building a system equation and a measurement equation are established, parameters are estimated in real time by using an exponential weighting average algorithm and Allan coefficient of variation is effectively estimated by using nonlinear adaptive kalman filtering. The method for eliminating noise on line is capable of updating error items of an optical fiber compass and improving the course and posture output accuracy of the compass.

The invention discloses a power grid information operation and maintenance monitoring method based on deep learning. The method is based on time series data information in a power grid information operation and maintenance monitoring system. Cleaned time series data are obtained through a proper data preprocessing technology; a prediction function of to-be-detected time sequence data is realized by utilizing a long-short-term memory neural network, so that a normal behavior model of a to-be-detected time sequence is constructed, and whether the to-be-detected time sequence has an abnormal phenomenon is further judged through a control chart based on exponential weighted moving average. The method faces any abnormity influenced by time in the field of power grid information operation and maintenance monitoring, has certain universality, and has very important scientific significance and application value for instructive processing after abnormity discovery and prevention of serious faults possibly caused by the abnormity.

The invention belongs to the field of switching operation synchronous control of four-break intelligent circuit breakers, specifically a signal transmission clock synchronization compensation method for a four-break series circuit breaker. The method comprises the steps: obtaining corresponding normalized forwarding queuing delay and normalized reverse queuing delay according to the timestamps ofeach main clock and each slave clock of a current message; obtaining an instant asymmetric offset of the corresponding clock according to the normalized forwarding queuing delay and normalized reversequeuing delay of the current message; enabling the clock instant asymmetric offsets of the current message to pass through an exponential weighting moving average filter, and obtaining a mean value of the corresponding clock instant asymmetric offsets; obtaining the clock offset of the current message through the mean value of the clock instant asymmetric offsets of the current message; and compensating for the signal transmission time of the four-break series circuit breaker according to the offset of each clock.

The invention provides a fractional-order observability analysis method for a pulsar navigation system. The method comprises a preparation stage and a system observability analysis stage; the preparation stage comprises establishing a spacecraft orbit kinetic model and a pulsar navigation model which are demanded by navigation filtering; and the system observability analysis stage comprises working out the fractional differential of time by a measurement module, obtaining a measurement matrix of fractional order, constructing an observability analysis matrix, rearranging the condition number, calculating the condition number of exponential weighting, so as to obtain the observability analysis result. The observability analysis result obtained according to the technical scheme of the invention is relatively accurate, past measurement information is fully utilized, influence of orbit elements on navigation precision can be embodied, calculation is simple and the method is convenient to achieve, system resource is saved and analysis results can be efficiently obtained.

The invention discloses a method for detecting tiny faults of a satellite attitude control system and based on locally linear embedding (LLE). The method is based on a locally linear embedding method,and comprises the steps that firstly, historical data is enabled to have a zero mean value and a unified variance, then the number of neighborhood points is determined by a dynamic locally linear embedding (DLLE) method, a weight matrix W is reconstructed dynamically, then low-dimensional embedding Y of a sample set is found through the obtained weight matrix W, and two statistics of T<2> and SPEare further acquired; then an exponentially weighted moving average (EWMA) is solved by using an EWMA algorithm, an EWMA statistic range of normal data is used as a threshold value for judging whether to-be-detected data has faults or not, a mapping matrix A is further obtained and used for calculating the statistics of T<2> and SPE of online data, and the corresponding EWMA statistics is calculated; and finally whether the EWMA statistic of the online data is greater than a control limit or not is judged, if the EWMA statistic is greater than the control limit, the system faults are generated, and if not, the system is normal. According to the method for detecting the tiny faults of the satellite attitude control system and based on LLE, deficiencies of an original algorithm are overcome, and the detecting performance of the algorithm towards the tiny faults is improved.

The invention discloses an intelligent vehicle prediction control method based on visual spatial-temporal characteristics. Firstly, a steering wheel angle prediction network is constructed, includinga spatial characteristic extraction network, N spatial-temporal characteristic extraction modules and a spatial-temporal characteristic map fusion prediction module, characteristic maps of different scales and different time steps are obtained by the spatial characteristic extraction network, the spatial-temporal characteristics are extracted from the characteristic map of each scale by the spatial-temporal characteristic extraction modules, then the spatial-temporal characteristic map fusion prediction module fuses the spatial-temporal characteristics of different scales to predict the steering wheel angle; and after the steering wheel angle prediction network is trained, a moment to be predicted is predicted, and exponential weighted average is performed on the predicted value of the steering wheel angle and the historical predicted value to obtain the final predicted value of the steering wheel angle. According to the method, the spatial-temporal information in the continuous imageframes can be effectively extracted, and the spatial-temporal information of different scales is fused together so that the prediction control precision of the intelligent vehicle is greatly improved.

The invention relates to a system and a method for multi-terminal automatic evaluation of enterprise safety management capability. The method comprises two large classes of indexes about machine-material and people-process-environment, wherein for the machine-material class of indexes, a weight is determined by historical data through a variable coefficient, and a total value is determined by iterating with a multiplier reinforcing process; for the people-process-environment class of indexes, a weight is determined by a target optimizing matrix, a total value is determined by exponential weighting multi-stage fuzz evaluation, a B / S evaluating system is built based on an algorithm, system terminals comprise a mobile phone, a computer, special evaluating equipment and the like, results are automatically evaluated and converted into a centesimal system by the system, and the results are sent to a user through a mail, a message, WeChat and QQ. Compared with the prior art, the system and the method have the advantages that the use of different algorithms for the machine-material class of indexes and the people-process-environment class of indexes is favorable for scientific and just evaluation, abnormal values are highlighted by the algorithm for the machine-material class of indexes, and the algorithm for the people-process-environment class of indexes reflects opinions from all parties and weakens abnormal value effects; system automatic calculation and multi-terminal evaluation push are adopted in evaluation, and the defects of difficult evaluation, difficult calculation and the like can be overcome.

The invention discloses a method and system for cleaning power transmission and transformation reliability evaluation big data, and the method comprises the steps: clustering of to-be-tested sample data through a hierarchical clustering CURE algorithm, and obtaining a clustering cluster of normal sample data; selecting boundary sample data based on the clustering cluster of the normal sample data;according to an abnormal data identification algorithm based on a boundary sample, identifying abnormal data in the to-be-tested sample data, and identifying the abnormal data; labeling the positionof the identified abnormal data in the to-be-tested sample data; correcting the abnormal data by applying an exponential weighting moving average method according to the marked position of the abnormal data to obtain a correction result of the abnormal data; and replacing abnormal data in the sample data with the correction result.

The invention discloses a Chernoff fusion method based on expectation maximization approximation, which comprises the steps of performing particle filtering on each sensor to obtain a local estimationresult, approximating the local estimation result into Gaussian mixture distribution by adopting an expectation maximization method at the same time, interacting a Gaussian mixture parameter among multiple sensors, then performing preliminary data fusion by using a Chernoff fusion method under a first-order approximation model, enabling the fusion result to act as an importance sampling function,recovering local particle samples of each sensor, calculating corresponding exponential weights at the same time, acquiring an exponential weighting result of each particle sample to act as a new particle sample, then approximating the new particle sample into Gaussian mixture distribution by using the expectation maximization method again, finally performing distributed data fusion according toa Chernoff fusion criterion, and calculating by using the fusion result to obtain an estimation state of the target. The method can achieve the optimal Chernoff fusion and acquire a high-precision conservative distributed data fusion result.

The invention discloses a traffic abnormity identification method and system based on travel time evolution, and a storage medium. The method includes: acquiring floating vehicle data, constructing atravel time distribution model and performing solving, then calculating an exponential weighted moving mean value and generating an exponential weighted moving mean value control chart, and finally conducting traffic abnormity identification through the exponential weighted moving mean value control chart. According to the invention, the floating vehicle data is fully utilized, meanwhile, the travel time distribution model is constructed and the exponential weighted moving mean value is calculated, and then the exponential weighted moving mean control chart is generated, so that a dynamic evolution process of the traffic state of a single road section or the whole system can be identified, an abnormal traffic event in a traffic network can be effectively identified, and the identificationaccuracy is greatly improved. The method can be widely applied to the traffic field.

The invention relates to the technical field of image registration. According to the improved SIFT-Delaunay-based heterogeneous image registration method, PPB filtering is performed on an SAR image, which is an airspace filtering algorithm, so that coherent speckle noise generated by the SAR image can be removed, and edge characteristics can be effectively maintained. Then, a method the same as that of a traditional SIFT algorithm is adopted during feature point extraction, an improved exponential weighted mean ratio operator is adopted for feature point gradient calculation, and high-precision edge strength information is obtained. And finally, rough registration is performed by adopting an Euclidean distance in the registration process of the feature points, fine registration of the feature points is performed by adopting a Delaunay triangular mesh method, and the Delaunay triangular mesh has uniqueness and has relatively good invariance to translation, rotation and geometric distortion of the image.

The invention relates to an environmental data restoration / filling method and system. The system comprises the following steps: drawing box diagrams of indoor and outdoor environment data with different attributes based on ORIGIN software; determining an abnormal data set according to the box diagram; deleting continuous abnormal or missing multi-attribute data in the abnormal data set by adoptinga tuple ignoring method to obtain a processed abnormal data set; repairing or filling continuous abnormal or missing single attribute data by adopting a distance weighted KNN algorithm; segmenting the processed abnormal data set; repairing or filling discontinuous abnormal or missing data in the segmented abnormal data set by adopting an improved exponential weighted moving average model; and integrating the repaired or filled data. According to the method, the problems of data missing, data exception and the like in different conditions in indoor and outdoor environment monitoring data in abig data platform can be systematically solved in a classified mode, and the method and the system have important significance in improving the data quality and guaranteeing the data mining quality and efficiency.

The invention relates to a symptom correlation early warning algorithm based on exponential weighted moving average, and belongs to the field of big data analysis. Specifically, a symptom incidence database is established, the collected symptom information is screened and processed and the symptom which is not in the database is eliminated according to the requirement of the familiar symptom. Theconventional correlation algorithm only calculates the basic average number of symptoms and ignores the influence of the time before and after the symptom data. According to the symptom correlation early warning algorithm based on the exponential weighted moving average, the mean obtained by exponential weighting is taken as the basis of the correlation algorithm so that the historical retrospective data and the current data are combined to make the correlation coefficient more accurate, then data comparison is performed according to the obtained correlation coefficient to find out the data abnormity and obtain the time point of early warning and the data source early warning and thus the better early warning effect can be achieved.

The fluctuation of prices of a stock market has inherent randomness; and meanwhile, the change of prices also reflects trending useful information. How to eliminate random noises from the change of market prices and extract the useful information becomes the problem needing to be urgently solved in investment. The invention provides a quantified financial investment system capable of eliminating the noises and a realization method therefor. The quantified financial investment system comprises a financial quotation interface module, an industry classification module, an index weighting module, an investment decision-making module and a core computing module, belonging to the field of financial investment. The fluctuation of prices of the stock market has inherent randomness; and meanwhile, the change of prices also reflects the trending useful information. According to the quantified financial investment system capable of eliminating the noises and the realization method therefor, stochastically disturbed and trending price change information is separated in two dimensions of space and time through analyzing a moving average price, and trending effective information in a stock industry is effectively extracted, so as to guide to improve the investment income.

The invention discloses a criminal and loan lost person multi-network joint search method based on mobile social network relationship closeness. The method comprises the following steps: constructinga mobile social network of a lost person; constructing a lost person interaction attribute matrix; constructing an exponential weight calculation model, and constructing an exponential weight interaction attribute matrix; establishing a distance calculation formula between any two nodes in the family members of the lost person; firstly, a maximum interaction user is defined, then a step-by-step elimination algorithm is proposed to determine a minimum interaction user of family members, and a relationship closeness calculation model between a lost person and any family member is established; alost person family member multi-network joint search algorithm and a lost person family member sequence search algorithm based on mobile social network relationship closeness are proposed to search lost persons. The method has practical help and operability for tracking escapers, tracing fund-carrying escapers by banks and tracking loaning lost persons by financial platforms, including lost persons found in society and families.