JP6841774B2 - Predictors, prediction methods and computer programs - Google Patents

Description

The present invention relates to predictors, predictors and computer programs.

Advances in network technology and sensing technology have made it possible to acquire diverse and enormous amounts of data at each sampling time. As one of the data that can be acquired for each sampling time, there is multidimensional time series data (hereinafter referred to as "spatial information data") including position information in space. Spatial information data includes, for example, time-series data including position information of a mobile terminal as a sensing device and statistical information of a user of the mobile terminal, vehicle position information of a car navigation system as a sensing device, and vehicle user's. It is time-series data including statistical information, time-series data including the position information of the ticket gate passed by the user of the transportation IC card and the statistical information of the user of the transportation IC card.

Spatial information data is expected to be used for marketing, infrastructure development, urban development, disaster support, etc. When using in these fields, spatial information data is analyzed according to the usage pattern. In order for the analysis of spatial information data to be performed with high accuracy, it is necessary to acquire and accumulate a large amount of spatial information data. On the other hand, an increase in the amount of spatial information data to be acquired leads to an increase in the cost required for accumulating the spatial information data. Therefore, the accumulated spatial information data needs to be efficiently compressed.

A data compression device, which is a device for compressing data, can utilize the correlation in the time direction in the spatial information data in order to efficiently compress the stored spatial information data. The correlation in the time direction is expressed by the flow rate.

B. Lucas and T. Kanade, “An iterative image registration technique with an application to stereovision”, In Proc. Seventh International Conference on Artificial Intelligence, pp.674-679, 1981.

However, in the spatial information data, there is a problem that information that can trace the flow rate is not given from the viewpoint of protecting the privacy of the observed person. Spatial information data to which information that can trace the flow rate is not given is represented as data representing independent frequencies for each time. Spatial information data is multidimensional time-series data because it is acquired at sampling time intervals. Therefore, the spatial information data is data representing the frequency for each spatial region acquired at the sampling time interval, and is independent multidimensional time series data for each time.

In view of the above circumstances, an object of the present invention is to provide a prediction device, a prediction method, and a computer program for estimating the correlation of time series data in the time direction.

One aspect of the present invention is used to determine a displacement vector sequence representing a prediction of the amount of change in the frequency at a future time of the time-series data based on time-series data including the frequency of the statistical value in the past time. A first displacement vector sequence generator that generates a plurality of first displacement vector sequences, and a first of the first predicted frequency and the frequency calculated based on the first displacement vector sequence for each of the first displacement vector sequences. A plurality of second displacement vector sequences representing candidates for the displacement vector sequence are generated based on the first difference calculation unit for calculating the difference and the first displacement vector sequence satisfying a predetermined condition among the first displacement vector sequences. The second difference between the second displacement vector sequence generator, the second predicted frequency calculated from the second displacement vector string, and the first predicted frequency is calculated, and the first difference and the second difference are obtained. A prediction device including a second difference calculation unit that determines the second displacement vector sequence as the displacement vector sequence when a predetermined condition is satisfied.

One aspect of the present invention is the above-mentioned prediction device, and the second displacement vector sequence generation unit is the second displacement vector when the first difference and the second difference do not satisfy a predetermined condition. Among the columns, a new second displacement vector sequence is generated based on the second displacement vector sequence that satisfies a predetermined condition.

One aspect of the present invention is the prediction device, wherein the first displacement vector sequence generation unit determines position information associated with a random number determined based on a difference between the time series data. The first displacement vector sequence is generated based on the position information.

One aspect of the present invention is the prediction device, in which the first displacement vector sequence generation unit generates the random numbers based on a predetermined probability.

One aspect of the present invention is the prediction device, and the second difference calculation unit is the first when the error between the minimum value of the first difference and the minimum value of the second difference is equal to or less than a threshold value. 2 The displacement vector sequence is determined as the displacement vector sequence.

One aspect of the present invention is the prediction device, wherein the second displacement vector sequence generation unit is based on the first displacement vector sequence in which the magnitude of the first difference is the lower 50%. 2 Generate a plurality of displacement vector sequences.

In one aspect of the present invention, the predictor represents a prediction of the amount of change in the frequency of the time-series data at a future time based on the time-series data including the frequency of the statistical value in the past time. A first displacement vector sequence generation step for generating a plurality of first displacement vector sequences used for determination, and a first prediction calculated by the prediction device for each of the first displacement vector sequences based on the first displacement vector sequence. The displacement vector sequence is based on the first difference calculation step for calculating the first difference between the frequency and the frequency, and the first displacement vector sequence in which the prediction device satisfies a predetermined condition among the first displacement vector sequences. A second displacement vector sequence generation step of generating a plurality of second displacement vector sequences representing the candidates of the above, and a second of the second predicted frequency and the first predicted frequency calculated from the second displacement vector sequence by the prediction device. A prediction method comprising a second difference calculation step of calculating a difference and determining the second displacement vector sequence as the displacement vector sequence when the first difference and the second difference satisfy a predetermined condition.

One aspect of the present invention is a computer program for operating a computer as the above-mentioned prediction device.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide a prediction device, a prediction method, and a computer program for estimating the correlation of time series data in the time direction.

It is a functional block diagram which shows the example of the functional structure of the estimation apparatus in embodiment. It is a flowchart which shows the example of the flow of the generation of the coordinate reference table in an embodiment. It is a flowchart which shows the example of the flow of calculation of the estimated value of the frequency in each mesh of a t + 1 frame in an embodiment. It is a flowchart which shows the flow of the process of generating the displacement vector sequence in embodiment. It is a flowchart which shows the flow of the time-direction correlation estimation processing in embodiment.

FIG. 1 is a functional block diagram showing an example of the functional configuration of the estimation device in the embodiment. The prediction device 100 is an information processing device that estimates time-series data at a future time based on the time-series data at the past time that constitutes the time-series data. The time series data includes, for example, the frequency or position information of the statistical value. For example, when the prediction device 100 is given the frequency F _q (t) at time t and the frequency F _q (t + 1) at _{time t + 1, the prediction device 100 is at time t that most reproduces the frequency F q} (t + 1). The flow rate is estimated by determining the displacement vector. The displacement vector represents the change in the frequency of the particles between frames. The displacement vector is a two-dimensional vector having a horizontal component and a vertical component. The displacement vector determines the position of each particle at the next time. The prediction device 100 notifies another device such as a data compression device of a flow rate signal representing the estimated flow rate.

The prediction device 100 includes a processor such as a CPU (Central Processing Unit) connected by a bus, a memory, an auxiliary storage device, and the like, and by executing a flow rate estimation program, the coordinate reference table generation unit 101 and the displacement vector sequence generation unit 102 , A device including a first prediction error calculation unit 103, a displacement vector discard unit 104, a displacement vector sequence regeneration unit 105, a displacement vector update unit 106, and a second prediction error calculation unit 107. All or part of each function of the prediction device 100 may be realized by using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The flow rate estimation program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a flexible disk, a magneto-optical disk, a portable medium such as a ROM or a CD-ROM, or a storage device such as a hard disk built in a computer system. The flow rate estimation program may be transmitted via a telecommunication line.

The prediction device 100 accepts a plurality of frames as time series data. For example, the prediction device 100 accepts a plurality of frames as spatial information data. Each frame is represented by a δ × δ mesh for each frame. Spatial information data may not have a flow rate that represents a temporal correlation. In this case, the spatial information data is data representing the frequency for each spatial region acquired at the sampling time interval. In the present embodiment, the frequency of the sample position (x, y) of the frame (hereinafter referred to as “t-th frame”) at time t is f (x, y, t). Each sample position (x, y) represents a δ × δ mesh. f (x, y, t) is the frequency contained in the mesh. Here, it is defined in the range of 0 ≦ x ≦ X-1, 0 ≦ y ≦ Y-1, and 0 ≦ t ≦ T. f (x, y, t) is quantized with the quantization width q, and the frequency after quantization (hereinafter referred to as “quantization frequency”) is expressed by the equation (1). It should be noted that f ₁ (x, y, t) = f (x, y, t).

The inside of the mesh at the sample position (x, y) of the t-frame is _{composed of f q} (x, y, t) particles. _{Therefore, the total number K q} of the particles at the sample position (x, y) in the t-frame is expressed by the following mathematical formula (2). Hereinafter, the entire particle F _q (t) of the t-frame is expressed by the mathematical formula (3).

The coordinate reference table generation unit 101 generates a coordinate reference table. Specifically, the coordinate reference table generation unit 101 assigns an index to the particles in the frame. The index is used to identify the particles within the frame. The index range is k = 0, 1, ... k _q -1. The coordinate reference table generation unit 101 unifies the mesh in the frame by raster scanning. The coordinate reference table generation unit 101 assigns particles in each mesh by assigning continuous indexes. The coordinate reference table generation unit 101 stores the spatial positions (x, y) corresponding to each index in the coordinate reference table. Each functional unit of the prediction device 100 refers to the coordinate reference table as needed.

The displacement vector sequence generation unit 102 generates a predetermined number of displacement vector sequences. The predetermined number is, for example, K _q . By generating the displacement vector sequence, the displacement vector sequence generation unit 102 generates the inter-frame difference calculation unit 121, the particle generation probability calculation unit 122, the cumulative particle generation probability calculation unit 123, the random number generation unit 124, and the displacement vector sequence determination unit 125. Functions as. The displacement vector sequence is a vector obtained by combining the _{displacement vectors V x, t} and V _{y, t.} V _{x, t} is an array consisting of _{K q elements.} The kth element V _{x, t} [k] (k = 0, ..., K _q -1) represents the horizontal displacement of the kth particle. V _{y and t} are arrays consisting of _{K q elements.} The kth element V _{y, t} [k] (k = 0, ..., K _q -1) represents the vertical displacement of the kth particle. The displacement vector sequence generation unit 102 is an aspect of the first displacement vector sequence generation unit. The first displacement vector sequence generation unit generates a plurality of first displacement vector sequences used for determining the displacement vector sequence at a future time based on the time series data. The displacement vector generated by the displacement vector sequence generation unit 102 is one aspect of the first displacement vector sequence. The first displacement vector sequence generator determines the position information associated with the random number determined based on the difference between the time series data, and uses the position information as the displacement vector of the kth particle to set the first displacement vector sequence. Generate.

The inter-frame difference calculation unit 121 calculates the inter-frame difference. The difference between frames is represented by d (x, y, t). d (x, y, t) is represented by the following mathematical formula (4).

The particle generation probability calculation unit 122 generates each element of each displacement vector sequence according to a random number based on the particle generation probability represented by the following mathematical formula (5).

Here, B in the mathematical formula (5) is a normalization constant represented by the mathematical formula (6).

Here, for (x + i, y + j) outside the range of 0 ≦ x + i ≦ X-1, 0 ≦ y + j ≦ Y-1, w (x + i, y + j) = 0. Further, g _σ (i, j) is a probability density function of a two-dimensional normal distribution. For g _σ (i, j), the mean is a zero vector, and the covariance matrix is expressed by the mathematical formula (7). The probability density function is represented by the mathematical formula (8).

The particle generation probability calculation unit 122 calculates d (x, y, t) ⁺ of the mathematical formula (5) based on the difference between the frames. The particle generation probability calculation unit 122 calculates d (x, y, t) ⁺ based on the mathematical formula (9).

The particle generation probability calculation unit 122 calculates the standard deviation parameter σ of the normal distribution of the equation (8) based on the difference between frames. The particle generation probability calculation unit 122 calculates σ based on the mathematical formula (10).

The variables α and β in the equation (10) are hyperparameters. The particle generation probability calculation unit 122 sets the hyperparameter β when f (x, y, t + 1) = 0. The variables α and β are predetermined. The particle generation probability calculation unit 122 calculates d (x, y, t) ⁻ of the mathematical formula (10) based on the difference between the frames. The particle generation probability calculation unit 122 calculates d (x, y, t) ⁻ based on the mathematical formula (11).

The particle generation probability calculation unit 122 calculates h (i, j) based on the displacement amount. The particle generation probability calculation unit 122 calculates h (i, j) based on the mathematical formula (12).

The cumulative particle generation probability calculation unit 123 calculates the cumulative particle generation probability. Specifically, the cumulative particle generation probability calculation unit 123 sets w (x + i, y + j) (i = -I, ..., I, j = -J, ..., J) as a unified series ~ w (n). ). The cumulative particle generation probability calculation unit 123 converts the unit series to w (n) by a known method such as raster scanning. The cumulative particle generation probability calculation unit 123 calculates the cumulative particle generation probability based on the mathematical formula (13) for the converted unit series ~ w (n).

The random number generation unit 124 generates a predetermined number of uniform random number vectors u according to a uniform distribution having [0,1] as a range. The predetermined number is the number of particles in the mesh. The random number generation unit 124 identifies n satisfying the mathematical formula (14) for each uniform random number vector u. The random number generation unit 124 determines n as a random number according to w (x + i, y + j). The random number generation unit 124 generates K _q random numbers.

The displacement vector sequence determination unit 125 calculates the two-dimensional coordinates for each random number based on the coordinate reference table. The displacement vector sequence determination unit 125 determines the calculated two-dimensional coordinates as the displacement vector of the kth particle. For example, in the displacement vector sequence determination unit 125, when the k-th generated random number is n, the corresponding two-dimensional coordinates are represented by an integer not exceeding n / (2J + 1). Here, n% (2I + 1) is an operation that returns the remainder when n is divided by 2I + 1.

The first prediction error calculation unit 103 calculates an estimated value ~ f _q (x, y, t + 1) of the frequency of each mesh in the t + 1 frame. The array having the estimated value ~ f _q (x, y, t + 1) as an element is represented by the mathematical formula (15) using _{the displacement vectors V x, t} , V _{y, t.} The array represented by the equation (15) represents the predicted frequency of the t + 1th frame.

The first prediction error calculation unit 103 calculates an error between the frequency of the t + 1 frame and the prediction frequency of the t + 1 frame (prediction error of the t + 1 frame). The prediction device 100 identifies the displacement vector sequence that minimizes the prediction error of the t + 1 frame as the prediction frequency of the t + 1 frame as the displacement vector sequence V ^* _{x, t} , V ^* _{y, t.} V ^* _{x, t} , V ^* _{y, t} are expressed by the mathematical formula (16). The first prediction error calculation unit 103 is an aspect of the first difference calculation unit. The first difference calculation unit calculates the first difference between the first predicted frequency calculated based on the first displacement vector sequence and the frequency included in the time series data for each first displacement vector sequence. The prediction error of the first t + 1 frame is one aspect of the first difference.

The displacement vector discarding unit 104 discards a displacement vector sequence that satisfies a predetermined condition. The predetermined condition is that, for example, the displacement vector discarding unit 104 sorts the calculated prediction error in descending order and sets it as the lower 50% of the sorted prediction error. In this case, the displacement vector discarding unit 104 discards the lower 50% displacement vector sequence.

The displacement vector sequence regeneration unit 105 randomly selects two displacement vector sequences according to a uniform random number from the displacement vector sequences remaining without being discarded. If the two selected displacement vector sequences are the same displacement vector sequence, the displacement vector sequence regeneration unit 105 selects the two displacement vector sequences again. The displacement vector sequence regeneration unit 105 generates two new displacement vectors by rearranging the particles between the two displacement vector sequences. In the displacement vector sequence regenerating unit 105, the sum of the displacement vector strings remaining without being discarded and the newly generated displacement vector string becomes equal to the sum of the displacement vector strings generated by the displacement vector string generating unit 102. To generate a new displacement vector sequence. The displacement vector sequence regeneration unit 105 is an aspect of the second displacement vector sequence generation unit. The second displacement vector string generation unit is based on the first displacement vector string that satisfies a predetermined condition (for example, the first displacement vector string in which the magnitude of the first difference is the lower 50%) among the first displacement vector strings. A plurality of second displacement vector sequences representing candidates for the displacement vector sequence are generated. The displacement vector sequence generated by the displacement vector sequence regeneration unit 105 is one aspect of the second displacement vector sequence.

The displacement vector update unit 106 selects particles in the displacement vector in the displacement vector sequence based on a predetermined probability. The displacement vector update unit 106 updates the displacement vector of the selected particle based on a random number. The predetermined probability is a parameter given from the outside of the prediction device 100.

The second prediction error calculation unit 107 calculates the prediction error of each displacement vector sequence. The second prediction error calculation unit 107 compares the obtained minimum value of the prediction error with the minimum value of the prediction error calculated by the first prediction error calculation unit 103. The second prediction error calculation unit 107 ends the process when the difference between the two minimum values is equal to or less than a predetermined threshold value. If it is larger than a predetermined threshold value, the displacement vector sequence updated by the displacement vector updating unit 106 is output to the displacement vector discarding unit 104. The second prediction error calculation unit 107 is an aspect of the second difference calculation unit. The second difference calculation unit calculates the second difference between the second predicted frequency and the first predicted frequency calculated from the second displacement vector sequence, and the first difference and the second difference are predetermined conditions (for example, the first difference). When the error between the minimum value of 1 difference and the minimum value of the second difference is less than or equal to the threshold value), the second displacement vector sequence is determined as the displacement vector sequence representing the amount of change in the frequency included in the time series data. ..

FIG. 2 is a flowchart showing an example of the flow of generating the coordinate reference table in the embodiment. The coordinate reference table generation unit 101 sets 0 to the variable k representing the elements of the array (step S101). The coordinate reference table generation unit 101 determines the values of the coordinate reference table in the order of y = 0, ..., Y-1 (step S102). The coordinate reference table generation unit 101 determines the values of the coordinate reference table in the order of x = 0, ..., X-1 (step S103).

The coordinate reference table generation unit 101 sets _x in Ψ x [k: k + f _q (x, y, t) -1] (step S104). Ψ _x [k: k + f _q (x, y, t) -1] represents each element from the kth element of _{Ψ x to the k + f q} (x, y, t) -1. The coordinate reference table generation unit 101 sets _y in Ψ y [k: k + f _q (x, y, t) -1] (step S105). Ψ _y [k: k + f _q (x, y, t) -1] represents each element from the kth element of _{Ψ y to the k + f q} (x, y, t) -1. The coordinate reference table generation unit 101 sets k + f _q (x, y, t) in k (step S106). When the value of the coordinate reference table is determined in all x (step S107) and the value of the coordinate reference table is determined in all y (step S108), the process of generating the coordinate reference table ends.

FIG. 3 is a flowchart showing an example of the flow of calculating the estimated value of the frequency in each mesh of the t + 1 frame in the embodiment. The first prediction error calculation unit 103 calculates the estimated value in the order of _{k = 0, ..., K q -1 (step S201).} The first prediction error calculation unit 103 determines whether or not the values _{of the horizontal displacement V x, t} [k] of the k-th particle of the t-frame are 0 and the values of the vertical displacement V _{y, t [k] are 0.} Is determined (step S202). When the values of the horizontal displacement V _{x, t} [k] are 0 and the values of the vertical displacement V _{y, t} [k] are 0 (step S202: YES), the first prediction error calculation unit 103 performs ^ f _q (Ψ). _{The value of f q} (Ψ _x [k], Ψ _y [k], t) is set in x [k], Ψ _y [k], t + 1), and the process proceeds to step S206 (step S203).

When the values of the horizontal displacement V _{x, t} [k] are 0 and the values of the vertical displacement V _{y, t} [k] are not 0 (step S202: NO), the first prediction error calculation unit 103 performs ^ f _q (Ψ _{x). The value of f q} (Ψ _x [k], Ψ _y [k], t) -1 is set in [k], Ψ _y [k], t + 1) (step S204). The first prediction error calculation unit 103 has ^ f _q (Ψ _x [k] + V _x [k], Ψ _y [k] + V _y [k], t + 1) and f (Ψ _x [k] + V _x [k], Set the value of Ψ _y [k] + V _y [k], t) + 1 (step S205). When the first prediction error calculation unit 103 calculates the estimated value of the frequency in all k (step S206), the first prediction error calculation unit 103 ends the calculation of the estimated value of the frequency in each mesh of the t + 1 frame.

FIG. 4 is a flowchart showing a flow of processing for generating a displacement vector sequence in the embodiment. The prediction device 100 is given in advance a frequency f (x, y, t) at the position (x, y) of the t-th frame and a frequency f (x, y, t + 1) at the position (x, y) of the t + 1 frame. Be done. The inter-frame difference calculation unit 121 calculates the inter-frame difference based on the mathematical formula (4) (step S301). The particle generation probability calculation unit 122 sets the variable n to 0 (step S302).

The particle generation probability calculation unit 122 determines the value of the particle generation probability in the order of y = 0, ..., Y-1 (step S303). The particle generation probability calculation unit 122 determines the value of the particle generation probability in the order of x = 0, ..., X-1 (step S304). The particle generation probability calculation unit 122 calculates ^{the value of d (x, y, t) −} based on the mathematical formula (11) (step S305). The particle generation probability calculation unit 122 calculates the value of the standard deviation parameter σ of the normal distribution based on the mathematical formula (10) (step S306). The hyperparameters α and β are predetermined. The particle generation probability calculation unit 122 calculates ^{the value of d (x, y, t) +} based on the mathematical formula (9) (step S307). The particle generation probability calculation unit 122 calculates the mutation amount particle generation probability w (x, y) based on the mathematical formula (5) (step S308). The particle generation probability calculation unit 122 calculates h (i, j) based on the mathematical formula (12).

The cumulative particle generation probability calculation unit 123 calculates the cumulative particle generation probability w (n) based on the mathematical formula (13) (step S309). The cumulative particle generation probability calculation unit 123 increments the variable n (step S310). When the particle generation probability calculation unit 122 determines the cumulative particle generation probability value in all x (step S311), and the particle generation probability calculation unit 122 determines the cumulative particle generation probability value in all y. (Step S312), the process of generating the displacement vector sequence transitions to step S313.

The displacement vector sequence determination unit 125 determines the displacement vector sequence in _{the order of k = 0, ..., K q} -1 (step S313). The random number generation unit 124 generates a uniform random number vector according to a uniform distribution having [0,1] as a range. (Step S314). The random number generation unit 124 generates a random number n satisfying the mathematical formula (14) based on the uniform random number vector (step S315). The displacement vector sequence determination unit 125 calculates the two-dimensional coordinates for the random numbers based on the coordinate reference table. The displacement vector sequence determination unit 125 determines the calculated two-dimensional coordinates as the displacement vector of the kth particle (step S316). When the displacement vector sequence is determined (step S317) in all k, the displacement vector sequence determination unit 125 ends the process of generating the displacement vector sequence.

FIG. 5 is a flowchart showing the flow of the correlation estimation process in the time direction in the embodiment. The displacement vector sequence generation unit 102 generates a displacement vector sequence based on the process shown in FIG. 4 (step S401). The first prediction error calculation unit 103 calculates an error between the frequency of the t + 1 frame and the prediction frequency of the t + 1 frame (prediction error of the t + 1 frame) (step S402). The predicted frequency of the t + 1th frame is calculated based on the process shown in FIG.

The displacement vector discarding unit 104 discards the displacement vector sequence that satisfies a predetermined condition among the generated displacement vector sequences (step S403). The predetermined condition may be, for example, the lower 50% of the sorted prediction error. The displacement vector sequence regeneration unit 105 generates a new displacement vector from the displacement vector sequence remaining without being discarded (step S404). The displacement vector updating unit 106 selects particles in the displacement vector based on the probability given from the outside, and updates the displacement vector of the selected particles (step S405). The second prediction error calculation unit 107 calculates the prediction error of the updated displacement vector sequence (step S406).

The second prediction error calculation unit 107 determines whether or not the difference between the minimum value of the obtained prediction error and the minimum value of the prediction error calculated by the first prediction error calculation unit 103 is equal to or less than a predetermined threshold value. (Step S407). When it is equal to or less than a predetermined threshold value (step S407: YES), the flow rate estimation process ends. If it is not less than or equal to a predetermined threshold (step S407: NO), the process proceeds to step S403.

As described above, in the prediction device 100 of the embodiment, the displacement vector sequence generation unit 102 generates the displacement vector sequence based on the inter-frame difference between the tth frame and the t + 1th frame. The first prediction error calculation unit 103, the displacement vector discard unit 104, the displacement vector sequence regeneration unit 105, the displacement vector update unit 106, and the second prediction error calculation unit 107 of the prediction device 100 are the displacement vector sequences generated last time. When the difference between the minimum value of the prediction error and the minimum value of the prediction error of the displacement vector string generated this time is equal to or less than a predetermined threshold value, the displacement vector string generated this time is determined as the estimated flow rate. Therefore, the prediction device 100 of the embodiment can estimate the time-series correlation of the time-series data, which is the unobserved information, even with respect to the time-series spatial information data in which only the frequency is observed.

Each function of the prediction device 100 in the above-described embodiment may be realized by a computer. In that case, the program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read by the computer system and executed. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, or a storage device such as a hard disk built in a computer system. Further, a "computer-readable recording medium" is a communication line for transmitting a program via a network such as the Internet or a communication line such as a telephone line, and dynamically holds the program for a short period of time. It may also include a program that holds a program for a certain period of time, such as a volatile memory inside a computer system that serves as a server or a client in that case. Further, the above program may be for realizing a part of the above-mentioned functions, and may be further realized for realizing the above-mentioned functions in combination with a program already recorded in the computer system. It may be realized by using a programmable logic device such as FPGA (Field Programmable Gate Array).

Although the embodiments of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs and the like within a range that does not deviate from the gist of the present invention.

The present invention is applicable to an estimation device that estimates the flow rate used to compress spatial information data.

100 ... Predictor, 101 ... Coordinate reference table generation unit, 102 ... Displacement vector sequence generation unit, 103 ... First prediction error calculation unit, 104 ... Displacement vector discard unit, 105 ... Displacement vector sequence regeneration unit, 106 ... Displacement vector Update unit, 107 ... 2nd prediction error calculation unit, 121 ... Interframe difference calculation unit, 122 ... Particle generation probability calculation unit, 123 ... Cumulative particle generation probability calculation unit, 124 ... Random generation unit, 125 ... Displacement vector sequence determination unit

Claims (8)

A plurality of first displacement vector sequences used to determine a displacement vector sequence representing a prediction of the amount of change in the frequency at a future time of the time series data based on time series data including the frequency of the statistical value in the past time. The first displacement vector sequence generator to be generated and
For each of the first displacement vector sequences, a first difference calculation unit that calculates a first difference between the first predicted frequency and the frequency calculated based on the first displacement vector sequence, and
A second displacement vector string generation unit that generates a plurality of second displacement vector sequences representing candidates for the displacement vector sequence based on the first displacement vector sequence that satisfies a predetermined condition among the first displacement vector sequences.
The second difference between the second predicted frequency and the first predicted frequency calculated from the second displacement vector sequence is calculated, and when the first difference and the second difference satisfy a predetermined condition, the second difference is obtained. A second difference calculation unit that determines the displacement vector sequence as the displacement vector sequence, and
A predictor. When the first difference and the second difference do not satisfy a predetermined condition, the second displacement vector sequence generation unit performs the second displacement vector sequence that satisfies the predetermined condition among the second displacement vector sequences. The predictor according to claim 1, wherein a new second displacement vector sequence is generated based on the above. The first displacement vector sequence generation unit determines the position information associated with the random number determined based on the difference between the time series data, and generates the first displacement vector sequence based on the position information. The prediction device according to claim 1 or 2. The prediction device according to claim 3, wherein the first displacement vector sequence generation unit generates the random number based on a predetermined probability. The second difference calculation unit determines the second displacement vector sequence as the displacement vector sequence when the error between the minimum value of the first difference and the minimum value of the second difference is equal to or less than the threshold value. 4. The prediction device according to any one of 4. Any of claims 1 to 5, wherein the second displacement vector sequence generation unit generates a plurality of the second displacement vector sequences based on the first displacement vector sequence in which the magnitude of the first difference is the lower 50%. The prediction device according to item 1. The first displacement used by the predictor to determine a displacement vector sequence that represents a prediction of the amount of change in the frequency of the time series data at a future time based on the time series data including the frequency of the statistical value at the past time. The first displacement vector sequence generation step to generate multiple vector sequences, and
A first difference calculation step in which the prediction device calculates the first difference between the first predicted frequency and the frequency calculated based on the first displacement vector sequence for each first displacement vector sequence.
A second displacement vector string generation in which the prediction device generates a plurality of second displacement vector sequences representing candidates for the displacement vector sequence based on the first displacement vector sequence satisfying a predetermined condition among the first displacement vector sequences. Steps and
When the prediction device calculates the second difference between the second prediction frequency calculated from the second displacement vector sequence and the first prediction frequency, and the first difference and the second difference satisfy a predetermined condition. , The second difference calculation step of determining the second displacement vector sequence as the displacement vector sequence,
Prediction method. A computer program for operating a computer as the prediction device according to any one of claims 1 to 6.

Images