Time sequence data fitting and compressing method
A technology of time series data and compression method, applied in the direction of electrical components, code conversion, etc., can solve the problems of real-time performance of compression and compression
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0033] figure 1 Shows the implementation process of the time series data fitting and compression method provided in the first embodiment of the present invention. This embodiment can be adapted to the compression of any dimensional time series data, and can perform time series linear fitting and time series nonlinear fitting. , Without loss of generality, detailed as follows:
[0034] In step S101, each component of the D-dimensional time series data at time t is represented by the sum of the linear combination of M time basis functions and the fitting error of the component.
[0035] In this embodiment, the D-dimensional time series data at time t {x 1 (t),x 2 (t),...,x D (t)}, each component of t=0,1,2,...,N-1 is represented by the sum of the linear combination of M time basis functions and the fitting error of the component, namely:
[0036] x 1 ( t ) = X k = 0 M - 1 w 1 , k f k ( t ) + e 1 ( t ) ...
Embodiment 2
[0054] figure 2 It shows the implementation process of the time series data fitting and compression method provided in the second embodiment of the present invention. The first embodiment is a batch processing fitting method. The main disadvantages of this method are: 1) The inverse matrix needs to be solved; 2 ). When fitting time series data by segments, it is necessary to constantly test the number of data points fitted by each segment to determine the optimal weight coefficient matrix of the segment. These two shortcomings severely limit the real-time performance of time series data fitting. Therefore, it is necessary to transform the first embodiment into an online time series data fitting method to meet the real-time requirements. This embodiment has excellent real-time performance and can be applied. The data fitting for infinite time series is detailed as follows:
[0055] In step S201, each component of the D-dimensional time series data at time t is represented by the ...
Embodiment 3
[0081] image 3 Shows the implementation process of the time series data fitting and compression method provided in the third embodiment of the present invention. When multiple data points are fitted, although the first and second embodiments can ensure that the sum of squares of the fitting error is the smallest, they cannot guarantee The fitting error of all data points is relatively small. In fact, when the number of fitted data points is large, some points will have large fitting errors. Therefore, in order to improve the fitting accuracy of the data, this embodiment adopts a segmentation method to fit multiple data points. The basic principle of achieving lossy compression is: when the number of data points of the current fitting segment is greater than M When, the output is two items: 1) the number of data points for each fitting segment; 2) the optimal weight coefficient matrix for each fitting segment. Using these two data, the fitted data value of all points in the se...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


