61 results about "Time series processing" patented technology

An optical scanner has an oscillation-type optical scanning mechanism which includes a movable member which resonates at a predetermined resonance frequency, a drive signal generator which generates a drive signal for allowing the movable member to resonate at the resonance frequency, and an oscillation signal generator which generates an oscillation signal in response to a change of the radiation direction of the light beams. Here, the oscillation signal generator includes a dot clock generator which generates a dot clock which becomes the reference of time-series processing by setting a frequency of the oscillation signal as a reference frequency.

A data processing apparatus includes a plurality of image processing units which execute different processes on input image data and an encoding unit which encodes the image data. A generation unit selects an image processing unit to be used from among a plurality of image processing units based on requests from a plurality of output destinations to process image data of each frame input. The selected image processing unit processes the image data in time-series and the encoding unit encodes the processed image data. Thus, a plurality of image data corresponding to requests from a plurality of output destinations are generated.

An object of the present invention is to provide an ultrasonic urinary volume sensor capable of estimating a urinary volume in the bladder accurately corresponding to any particular individuals and / or conditions thereof by incorporating a space time series processing method allowing for an accurate estimation of the urinary volume in the bladder in association with the particular individuals and / or conditions thereof. An ultrasonic urinary volume sensor of the present invention comprises: a probe 1 having a plurality of ultrasonic oscillators 3 for oscillating ultrasonic waves toward a wall surface of the bladder, which is adhesively placed over a body surface in an abdominal section via an ultrasonic wave transmission medium interposed therebetween; and a processing section 2 for detecting and processing reflective echoes of the ultrasonic waves from the wall surface of the bladder, which have been oscillated by the plurality of ultrasonic oscillators 3 of the probe 1, wherein the plurality of ultrasonic oscillators 3 is disposed along a direction of expansion of the bladder.

A system and related device emulates the behavior of mobile networks independently from the manner of collecting data about the behavior of the network obtained through a simulation of field measurements. The system emulates the behavior of the network during the exchange of information between terminals and uses an emulation device comprising two circuit blocks. The first is able to process the information according to time sequences corresponding to those of mobile networks. The second one exchanges information between the terminals connected to suitable interfaces. This architecture makes it possible to emulate uplink and downlink functions of the mobile network, displaying on the terminals the effects of the behavior of the network under various operating conditions.

A data processing apparatus includes a plurality of image processing units which execute different processes on input image data and an encoding unit which encodes the image data. A generation unit selects an image processing unit to be used from among a plurality of image processing units based on requests from a plurality of output destinations to process image data of each frame input. The selected image processing unit processes the image data in time-series and the encoding unit encodes the processed image data. Thus, a plurality of image data corresponding to requests from a plurality of output destinations are generated.

The invention discloses an audio and video mutual retrieval method based on user click behaviors. The method comprises the following steps: preprocessing input audio and video data; Sending the preprocessed audio data into a deep convolutional neural network to obtain an audio representation vector and attention weight distribution; Sending the preprocessed video key frame into a deep convolutional neural network to obtain a key frame representation vector, and sequentially sending the key frame representation vector into a time sequence processing network based on an attention mechanism to obtain a representation vector of the video and attention weight distribution; Calculating the similarity of the audio and video representation vectors and sorting the audio and video according to the similarity; performing Annotating according to the attention weight distribution to provide explainable basis for sorting; Calculating the loss function through a user click behavior, and carrying outmodel training by adopting a backward propagation method; And carrying out retrieval matching on audios and videos in the media library based on the trained model. According to the method and the device, matched audios and videos in the media library can be retrieved under the condition of giving videos and audios.

The invention discloses a personnel flow monitoring method and apparatus based on a mobile network. The method comprises: acquiring time point position information of a mobile terminal and geographical position information of a base station; performing time sequence processing based on the position information of the mobile terminal at each time point, and obtaining the position information of the mobile terminal in each time period; matching the geographical position information of the base station with the preset geographical position information of a monitoring point, and obtaining a base station within the monitoring scope of the monitoring point; and analyzing the position information of the mobile terminal at each time point, the position information of the mobile terminal in each time period and the base station within the monitoring scope of the monitoring point, and obtaining the flow information of the mobile terminal within the monitoring scope of the monitoring point as personnel flow information. By using the method and apparatus provided by the invention, the problems of large resource investment and incapability of tracing back a personnel flow direction during conventional personnel flow monitoring can be solved.

There is provided a fall detector for use in detecting falls by a user, the fall detector comprising an accelerometer for producing a time series of measurement samples representing the acceleration acting on the fall detector; a processor for estimating a vertical velocity and / or vertical displacement of the fall detector from the measurement samples and using the estimated vertical velocity and / or vertical displacement to determine whether the user has suffered a fall; wherein the processor is configured to estimate a vertical velocity and / or vertical displacement of the fall detector from the measurement samples by estimating a corresponding time series of unit vectors representing acceleration due to gravity in the reference frame of the accelerometer from the time series of measurement samples; projecting each measurement sample onto the corresponding unit vector and subtracting acceleration due to gravity to give a series of estimates for the vertical acceleration of the fall detector; and integrating the series of estimates for the vertical acceleration over a time period to give a time series of values for the vertical velocity and / or vertical displacement of the fall detector.

The invention discloses a time series processing method and device. The time series processing method comprises the steps that a first time series is analyzed, and at least one second time series relevant to the first time series is obtained; multiple regression processing is conducted on the second time series, and a first function with the second time series as a dependent variable and a fitting series as an independent variable is obtained; according to the first function, the fitting series corresponding to the first time series is calculated; the difference between the fitting series and the first time series is detected; when the difference between the fitting series and the first time series is not larger than a threshold, a residual error series between the fitting series and the first time series is worked out; stationary processing is conducted on the residual error series, a stationary series corresponding to the residual error series is obtained, autoregressive moving average processing is conducted on the stationary series, and a second function for the relation between a front element and a rear element in the stationary series is obtained; according to the first function and the second function, a prediction result of a prediction time is obtained.

The invention discloses a multi-source data and time sequence processing method and device for construction of an industry knowledge graph. The method comprises the following steps: constructing an ontology layer of the knowledge graph, wherein the ontology layer comprises ontologies, ontology attributes and an ontology relation; extracting entities and entity attributes from multiple data sources, and conducting inconsistency check on the entities; conducting inconsistency check on the entity attributes of all the entities, wherein an entity relationship between the entities inherits the ontology relationship between the ontologies corresponding to the entities; and establishing indexes of a knowledge graph database and a time sequence database. The method has the advantages that throughthe uniqueness of a standard naming table and the relational database, entity conflict resolution accuracy is ensured to the greatest extent; by utilizing the advantage that the knowledge graph has the ontology layer, accurate classification of the entity attributes is realized by establishing similar labels, and fusion efficiency is effectively improved; and the problem that the knowledge graph is difficult to process time sequence data is solved by establishing indexes between knowledge graph database ontologies and a time sequence database form.

The invention provides a method for determining an undetermined frequent pattern in an undetermined time sequence, belonging to the field of time sequence processing. The method comprises the following steps of: S1, selecting the type of the undetermined frequent pattern and setting a time threshold value and a probability threshold value eta, wherein the type of the undetermined frequent pattern comprises a minimal occurrence pattern and a non-overlapping occurrence pattern; S2, calculating effective examples of each candidate pattern in the undetermined time sequence according to the type of the undetermined frequent pattern, and determining the type of each candidate pattern to be a first candidate pattern or a second candidate pattern; S3, judging whether the first candidate pattern is the undetermined frequent pattern by using a dynamic programming technology; and S4, judging whether the second candidate pattern is the undetermined frequent pattern by combining a state space compression coding technology and a fuzzy programming technology. According to the method disclosed by the invention, the undetermined frequent pattern in the undetermined time sequence can be accurately determined; and moreover, the calculation efficiency is improved by using the state space compression coding technology.

The invention discloses a moving point target detection method and system based on high-time-phase detection, and the method comprises the steps: obtaining an image sequence shot by a high-frame-frequency camera, and dividing the image sequence into a plurality of basic processing units; in each basic processing unit, detecting a pixel with a target by using time domain transient fluctuation information caused when the target passes through the pixel to obtain a track of the target; and associating the target tracks of all the basic processing units through a track association method to obtaina target overall track. According to the method provided by the invention, a high-frame-frequency detection data and time sequence processing method is utilized, so that the detection accuracy of thelow-signal-to-noise-ratio moving point target can be improved, and the false alarm rate is reduced.

The invention provides an object detector for a vehicle that can ensure sufficient S / N ratio and avoid error detection even if structure of a time sequence filter for receiving signal does not become complex. The detector has a L / D apparatus (3) that control light emitting of laser and control light receiving of reflector light, and an object detection part that detects object by receiving signal from the L / D apparatus (3). The object detection part includes: an accumulative data memorizer (43) that received signals are cumulatively added up for each of areas into which the target area is divided; a region measurement data storage memorizer (44) that measures data according with each region storage region; a time series processor (45) that noise is removed from obtained data by time series filtering; the object detector (47), if the value of a peak in data is less than that of the corresponding peak obtained in the previous scan by more than a specified value, such data are considered as an after image and not detected as an object.

The invention discloses an order-preserving sequence pattern mining method and relates to the technical field of electric digital data processing. According to the method, a candidate mode is generated by using a mode fusion method; the number of candidate modes is reduced, and the mode support degree of the candidate modes is calculated through a series of conversion and verification steps. The defects that in the prior art, for mining frequent patterns from a time sequence, solving accuracy, generality and completeness are difficult to consider at the same time, it is difficult to ensure that important information is not lost when the time sequence is processed, and it is difficult to analyze key trends by mining the frequent patterns are overcome.

The present invention provides a process and apparatus for double-face ink-printing canvases for advertising light boxes. The process includes: disposing a print head at a side of an ink print workbench, an unprinted face of a reverse side of the canvas facing to the print head; disposing an inductor at the other side of the ink print workbench, a printed face of an obverse side of the canvas facing to the inductor; collecting image data of the printed face by the inductor during conveying the canvas and transmitting the image data to a signal processing device to process, and after time sequence processing, determining print positions on the unprinted face depending on results of the processing; and driving the print head to print on the determined print positions on the unprinted face by a print driving device, for achieving the double-faced ink print on the canvas for the advertising light boxes.

The invention discloses a method and device for optimizing control parameters in a clinker firing system. The method comprises the steps of converting historical sample time series data into a historical parameter recommendation sample set according to a preset time series processing strategy, wherein each parameter in the historical parameter recommendation sample set is obtained by aggregating according to a production time sequence based on a mechanism characteristic and a lag period of a process link to which the parameter belongs; determining an alternative parameter recommendation sampleset based on the real-time parameter set and the historical parameter recommendation sample set of the clinker generation system; and transmitting each alternative recommendation sample in the alternative parameter recommendation sample set to a preset multi-target parameter optimization function for calculation, determining a target parameter recommendation sample based on a calculation result,and taking a target control parameter in the target parameter recommendation sample as a control parameter of the clinker firing system. According to the process, the parameters in the alternative parameter recommendation sample set take the lag period into consideration, time lag of associated data before and after is avoided, and the accuracy of target control parameter determination is guaranteed.

The invention discloses a time series unsupervised anomaly detection method based on a conditional regularization flow model. The time series unsupervised anomaly detection method comprises the following steps of: (1) preprocessing a time series, and constructing a training data set; (2) constructing a recurrent neural network for processing a historical time series into implicit representation; (3) constructing the conditional regularization flow model for modeling probability density of an observation window by taking a historical observation series as a condition, wherein the conditional regularization flow model is used for calculating conditional logarithm likelihood of a time series in the observation window; (4) learning optimization model parameters based on a maximum likelihood principle; (5) selecting a threshold value according to the conditional logarithm likelihood of all the samples under the conditional regularization flow model with determined parameters; (6) and calculating the conditional logarithm likelihood of the time series in the observation window on line by using the recurrent neural network and the conditional regularization flow model with the determinedparameters, and reporting the observation window to be abnormal when the conditional logarithm likelihood is lower than a specified threshold value. The time series unsupervised anomaly detection method can effectively reduce the false alarm rate of anomaly detection.

The invention provides a fault diagnosis method for stamping equipment, and belongs to the technical field of fault diagnosis of heavy equipment. The method comprises the following steps: acquiring basic data, current operation data and historical fault data of a hydraulic system of the stamping equipment, and transmitting and storing the data to terminal equipment; carrying out time-series processing on the current operation data; building a deep-learning network model based on an LSTM (Long Short-Term Memory) and a RNN (Recurrent Neural Network); training a network model by adopting an improved particle swarm POS algorithm according to the basic data and the historical fault data, optimizing parameters of the network model, and acquiring an oil leakage feature set; inputting the preprocessed current operation data into the network model to extract performance state features of the hydraulic system; and inputting the performance state features into a classifier to judge whether hydraulic equipment has an oil leakage risk or not. The method can quickly and accurately forecast whether the stamping equipment has the oil leakage risk or not, guarantee safe and reliable operation of the stamping equipment and reduce the operation and maintenance cost of the stamping equipment.

The invention relates to a wireless communication system for a charging pile. The system comprises the charging pile, a wireless module and a server, wherein the wireless module comprises a main control unit as well as a 2G wireless chip, an SD card unit and a 485 chip connected with the main control unit, and the wireless module transmits data with the charging pile and the server via the 485 chip and the 2G wireless chip respectively; the main control unit comprises an FAT file system unit, a message queue processing unit, a data uploading and transmission unit and a link detection unit communicating with each other in sequence, the message queue processing unit is connected with a logic time series processing unit, and the main control unit is connected with the SD card via the FAT filesystem unit and connected with the 2G wireless chip via the link detection unit. The system can ensure the integrity of data interaction between the charging pile and the remote server, and is smoother in data transmission and low in processing difficulty of an application layer of the server.

The invention discloses a temperature based power load data long-term prediction method. The method comprises the following steps: acquiring power load data and daily mean temperature of collection time points in a certain time interval; and with the daily mean temperature as a description index of power load data and based on the acquired power load data and a regression method based marginal increment model, predicting the daily power load value at a certain daily mean temperature one day. According to the method disclosed by the invention, temperature is used as the description index of power load data, so that the prediction of the power load data is effectively related to temperature, and moreover, the primitiveness of data is not damaged in processing. In addition, aiming at the defect of data set deficiency, a marginal increment is set for making up for the deficiency; in a traditional method in which load is converted into a time sequence to be processed, medium and long-term prediction cannot be realized; and the technical scheme disclosed by the invention can well solve the problem.

A building management system (BMS) includes building equipment configured to provide raw data samples of one or more data points in the BMS. The BMS further includes a data collector configured to collect raw data samples from the building equipment and generate one or more raw data time series comprising a plurality of the raw data samples. The BMS also includes a time series processing engine. The time series processing engine is configured to identify one or more time series processing workflows that apply to the raw data time series, each of the workflows comprising a predefined sequence of time series processing operation. The time series processing engine is further configured to process the raw data time series using the identified time series processing workflows to generate one ormore derived data time series. The BMS further includes a time series storage interface configured to store the raw data time series and the derived data time series in a time series database.

The invention provides a method and an apparatus for constructing a computational flow graph of time series processing, wherein, the method comprises the following steps of: acquiring a time series processing algorithm to be processed; Splitting the time series processing algorithm to be processed into a plurality of calculation expressions; Constructing a computational flow graph for the plurality of computational expressions to obtain a plurality of computational flow graphs; The plurality of computational flow diagrams are combined to obtain a flow diagram model corresponding to the time series processing algorithm to be processed. As a result of that technical solution provide by the embodiments of the present application, the data dependency of time series algorithm can be visually displayed, this dependency can be used to optimize the computational steps in the time series processing algorithm, The problem of processor congestion caused by data dependency can be avoided, and thecomputational flow diagram can be easily divided according to the characteristics of the computing system to match the computing power of the computing system, so as to improve the resource utilization rate of the computing system and the running efficiency of the time series processing algorithm.

An embodiment of the invention provides a time series processing method and a time series processing device and relates to the technical field of data mining. The time series processing method includes acquiring a target parameter time series of target parameters including parameters to be detected within a preset time length; then acquiring an influence factor time series of the target parameters within the preset time duration; finally, selecting a dynamic similarity series of the parameters to be detected from M subseries of the target parameters based on the target parameter time series and the influence factor time series. The time series processing method has the advantage that change processes of the target parameters and all influence factors thereof are included into selection rules of similar time series comprehensively, and similar time series selection accuracy and time series predication accuracy are both higher in time series prediction affected by accumulative effects, so that prediction errors of the time series in the prior art are reduced effectively.

The invention relates to a stock index prediction method and device based on a neural network model and a time sequence, and the method comprises the steps: extracting time sequence data of a to-be-predicted stock index, and obtaining feature data; preprocessing the feature data to obtain first data; performing time sequence processing on the first data to obtain second data; and training the second data through a neural network model, and predicting the price of the stock finger to be predicted, thereby improving the accuracy and length of stock finger prediction.

The invention discloses a time series data processing method, device and system, a server and a readable storage medium. The method comprises the following steps: receiving a plurality of pieces of to-be-processed data sent by a plurality of pieces of equipment, wherein the plurality of pieces of to-be-processed data respectively comprise time series data acquired by the plurality of pieces of equipment and identification codes of the plurality of pieces of equipment; according to grouping results of the multiple pieces of equipment, respectively importing multiple pieces of to-be-processed data into corresponding task queues respectively for sequential processing, and correspondingly obtaining multiple pieces of result data; and sequentially storing the result data into a first cache queue. According to the time series data processing method provided by the invention, based on the grouping result of the equipment, high-performance time series processing of large-batch data can be realized in parallel, and the advantages of multi-core operation are exerted. Meanwhile, a single machine can process data of tens of thousands of devices, a complex computer group is prevented from beingbuilt, and the deployment cost is remarkably reduced.

The invention relates to the field of landslide disaster prediction, and discloses an earth surface deformation field anomaly detection method based on a neural network. The method is based on a supervised classification method using a combination of sight multi-time and multi-geometric interferometric synthetic aperture radar (InSAR) time sequence displacement. A training data set is created that has been compared to independently validated data and currently most advanced classification techniques. According to the method, a full-automatic anomaly detection system using an InSAR surface deformation time sequence is created. Compared with InSAR time sequence processing, the method enables the processing time after abnormity identification to be negligible. By displaying how the proposed neural network system provides an accurate and temporarily feasible solution, compared with a radio frequency method and an analysis model, the abnormal time sequence detection processing time is greatly reduced.

The invention provides an interpretable recommendation model training method and device. The method comprises the following steps: obtaining a time sequence of user behaviors of multiple users; processing an initial knowledge graph based on the time sequence to obtain a time sequence knowledge graph; obtaining the feature vectors of the time sequence knowledge graph by embedding a joint learning model, wherein the feature vectors comprise entity feature vectors and relation feature vectors; and training an initial model according to the feature vectors to obtain a target recommendation model.

Embodiments of the present specification provide a time series processing method and device, and an electronic apparatus. When the method is implemented, a time series generated in the marketing campaign is provided; the time series includes at least two exception data. The method may include selecting a first representative anomaly data and a second representative anomaly data from the time series; calculating a first representative value of the first sub-time series and a second representative value of the second sub-time series, wherein the first sub-time series includes data that precedesthe first representative exception data within the time series and the second sub-time series includes data located within the time series after the second representative abnormal data; determining aneffect type of the marketing campaign based on the first representative value and the second representative value.

The invention discloses an unstructured grid significant wave height prediction system based on deep learning, which comprises a long short-term memory (LSTM) model, and the unstructured grid significant wave height prediction system comprises a time sequence processing model composed of two LSTMs and is a recurrent neural network with an input gate, a forget gate and an output gate. The input of the model is the effective wave height data of the unstructured grid in 10000 hours provided by the FVCOM mode, and the final output is the predicted wave height in the future 1 hour, 6 hours, 12 hours and 24 hours.