50 results about "Time-series segmentation" patented technology

A learning motion control is made by using a hierarchical recurrent neural network. A variety of motion patterns can be created by segmenting a motion pattern taught by a human into parts, automatically in time series, combining the parts, and controlling the motion of the body. Local time-series patterns and a perspective pattern combination of the local time-series patterns are formed by time-series segmentation. The control stability of the motion pattern is ensured since the ZMP stability criteria witch are the indices of the static and dynamic stabilities of body are satisfied.

The invention discloses an abnormal alarm data detection method based on a multivariate time series. The abnormal alarm data detection method based on the multivariate time series comprises the steps that data of multiple correlated variables are extracted from historical data, the multivariate time series is established and standardized, and the symbol direction between the variables in the normal state is calculated; time series segmentation description based on key turning points is determined, the minimum time interval is set, and key turning point searching is conducted; the piecewise linearity of the multivariate time series is represented, a fitting error is determined according to the orthorhombic distance between a data point and each segment, a loss function threshold value is set, the number of the segments is optimized, and an optimized segmentation result is obtained; and based on the optimized segmentation result, correlation analysis is conducted on all the segments of the multivariate time series, the symbol direction between the segment variables is extracted, and abnormal data with the symbol direction inconsistent with the symbol direction in the normal state are detected. By adoption of the abnormal alarm data detection method based on the multivariate time series, favorable conditions are provided for designing of a dynamic alarm threshold value of a multivariable alarm system, and thus disturbance alarms are reduced.

A computer-assisted method for quantitative characterization and treatment of ventricular fibrillation includes preprocessing a time series of an atrial fibrillation signal obtained from a patient, segmenting the time series of the AF signal into activation segments by the computer system, obtaining local activation waveforms (LAW) from the activation segments, determining degrees of similarity between the LAWs, and identifying one or more critical regions in the patient's atria if the LAWs have degrees of similarity exceeding a first threshold value.

A method for accelerating time series data base queries includes segmenting an original time series of signal values into non-overlapping chunks, where a time-scale for each of the chunks is much less than the time scale of the entire time series, representing time series signal values in each chunk as a weighted superposition of atoms that are members of a shape dictionary to create a compressed time series, storing the original time series and the compressed time series into a database, determining whether a query is answerable using the compressed time series or the original time series, and whether answering the query using the compressed time series is faster. If answering the query is faster using the compressed representation, the query is executed on weight coefficients of the compressed time series to produce a query result, and the query result is translated back into an uncompressed representation.

The invention provides an intersection signal control time period dividing method. The intersection signal control time period dividing method comprises the steps of: acquiring traffic flow data of anintersection in one day; converting the traffic flow data into traffic flow vectors, and acquiring a distance between every two adjacent traffic flow vectors; dividing a time period of a day into initial time segments based on a CUSUM time series segmentation algorithm based on big data; combining the initial time segments according to a preset manner to obtain combined time segments; adopting aclustering algorithm for classifying the combined time segments according to the distance between every two adjacent traffic flow vectors, so as to obtain the classified target time segment; and determining a signal control strategy of each segment according to the target time segment. The intersection signal control time period dividing method has low dependence on traffic information acquisition, has good reliability, can significantly improve the traffic passing capacity and effectively reduces the traffic delay.

A method and system are provided. The method includes extracting shapelets from each of a plurality of time series dimensions of multi-dimensional time series data. The method further includes building a plurality of decision-tree classifiers, one for each time series dimension, responsive to the shapelets extracted therefrom. The method also includes generating a pairwise similarity matrix between respective different ones of the plurality of time series dimensions using the shapelets as intermediaries for determining similarity. The method additionally includes applying a feature selection technique to the matrix to determine respective feature weights for each of shapelet features of the shapelets and respective classifier weights for each of the decision-tree classifiers that uses the shapelet features. The method further includes combining decisions issued from the decision-tree classifiers to generate a final verdict of classification for a time series dimension responsive to the respective feature weights and the respective classifier weights.

For each one note or for each plurality of notes constituting a reference tone, a segment setting section segments a time series of actual pitches of the reference tone into one or more note segments. For each of the one or more note segments, a relativization section creates a time series of relative pitches that are relative values of individual ones of the actual pitches of the reference tone to a normal pitch of the note of the note segment. Information registration section stores, into a storage device, relative pitch information comprising the time series of relative pitches of each individual one of the note segments. The segment setting section may use musical score data, time-serially designating the notes of the reference tone, to set each of the note segments for each note designated by the musical score data, and may correct at least one of start and end points of each of the set note segments in response to user's operation.

Data processing method for analysis of the energy consumption of a site on the basis of measurements including:selection of data time series,segmentation of the time series into segments,projection of the segments,display of the projections of the data segments,e) selection of at least one first group of data segments,establishment of a digital model of classification characterising the first group of data segments, selected on the basis of calendar metadataestablishment of a digital model of classification characterising the first group of data segments, selected on the basis of the segments' profiles (FIG. 5B).

The invention discloses a life prediction method based on the change point detection of an associated synchronous time series signal, which is used to solve the life prediction problem under the interactive influence of large time scale and multi-dimensional time series. First, the original multi-dimensional synchronous time series is resampled to obtain the equivalent time series signal after scale reduction; secondly, a likelihood function is constructed for the sampled signal with the change point vector, piecewise slope and piecewise variance as parameters ; Then, for different values ​​of the number of change points, the maximum likelihood estimation is used to estimate the best value of the change point vector, piecewise slope and piecewise variance; The change point vector corresponding to the selected optimal number of change points divides the original time series into a plurality of local time series that are not constant zero and have only one change point, and repeatedly construct the likelihood function and In the step of optimal change point estimation, the final change point position of the original time series is obtained; finally, the signal at which the final change point is located is determined, and the time series at the tail of each dimension signal is obtained, which is used for parameter estimation required for subsequent life prediction.

Provided is a behavior video non-supervision time-sequence partitioning method. The method comprises steps of: initializing the initial time of video detection as nt and a corresponding sliding window frame length as Lt; detecting behavior change points in an established video sequence window; if it is detected that a behavior change point c is in the video sequence window, using a time point c as the initial time of detection and reinitializing sliding window frame length in order to continue detecting subsequent videos; if it is not detected that the behavior change point is in the video sequence window, still using the nt as the initial frame of detection, namely nt+1=nt, and updating the sliding window frame length as Lt+1=Lt+[delta]L, wherein the [delta]L is a incremented step length of the length of the sliding window; ending the method until all video frame sequences are detected or reach a predetermined deadline T0. The method makes a decision on data change points in behavior video analysis, does not require online and real-time non-supervision partition with prior knowledge, and is directly used in behavior video data online analysis.

The invention provides a dynamic system gain estimation method based on a historical data steady-state value, and the method comprises the steps: firstly, dividing an input time sequence and an outputtime sequence into short data segments through employing a linear segmentation representation method, and finding out data segments with input and output being in a steady state at the same time; calculating input and output steady-state values from the data segment under the steady-state condition; elements, in the steady-state values, associated with the same steady-state gain in statistical significance are found, the elements are divided into one group, the steady-state gain of each group is estimated, and interval estimation of estimation parameters is given. According to the method, theinput and output steady-state values can be automatically found in the historical data sample, a plurality of steady-state gains under different working conditions can be accurately and effectively estimated, verification is carried out through a visualization method, and the problems that the time consumption for searching the steady-state values is long, the steady-state values are easily influenced by nonlinearity and the steady-state gain change is difficult to detect are solved.

The invention belongs to the technical field of displacement detection, and discloses a short-term displacement detection method based on GNSS dynamic positioning time series segmentation. First, the original coordinate time series is obtained through GNSS dynamic positioning; and then the detection window is used to perform time series segmentation to obtain a set of possible change points; For each possible change point, obtain the statistical value of the possible change point, and judge whether the possible change point is a change point according to the statistical value and the preset threshold; finally, for each change point, obtain the previous displacement estimation value corresponding to the change point, based on the previous Displacement Estimates Obtain the final displacement estimates. The invention solves the problem that the prior art cannot effectively detect and extract the displacement in a short time, and can effectively detect and extract the displacement in a short time.

The invention discloses a spatial time sequence data stream application-based Skyline query method, belongs to the field of application of a dynamic Skyline query in a data stream, and aims to solve the problem in real-time query processing of massive data. According to the technical key points, the method comprises the steps of S1, segmenting a continuous time sequence into a plurality of time slices according to time windows based on spatial time sequence division; S2, for each time slice, generating a grid inverted index; and S3, mapping moment query points to corresponding Skyline grids, then obtaining global Skyline grids by using a global Skyline grid calculation method to serve as a candidate set, performing the dynamic Skyline query on network node data in the candidate set according to a time sequence, and performing calculation to obtain an effective global Skyline result. The method has the effects that the result query is performed at a moment when executive time is over, so that the result is more accurate and conforms to the actual condition.

The invention provides a method for supporting large-scale time sequence data interaction based on a line segment KD tree, and the method comprises the steps of segmenting each time sequence into a plurality of line segments, and calculating the slope corresponding to each line segment; establishing a KD tree index structure in a three-dimensional space for all segmented line segments; calculating spatial span information in each KD tree node; and for the interaction target region, querying all line segment information passing through the selected region in the KD tree index structure, and carrying out interactive query. According to the invention, the problem of difficult query in current large-scale data analysis is effectively solved, and user observation is facilitated.