JP2015082800A - Residence position estimation device and residence position estimation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate an average residence position of a user in a resident state by appropriately discriminating a state under move/under residence on the basis of position data obtained by cell based or delay amount based positioning.SOLUTION: A residence position estimation device 10 includes: a preprocessing part 11 for providing processing data by giving a residence period expectation value and an arrival time expectation value of a location cell to location cell data including a terminal identifier, a positioning time, a location cell identifier and a delay amount; a group extraction part 12 in which a time sequential list is obtained by extracting a record for each terminal from the processing data and rearranging the records in the order of arrival time expectation values, the time sequential list is divided into residence cell list candidates of neighboring cells on the basis of a neighboring relation of cells and residence cell list candidates of which the sum of residence period expectation values is equal to or more than a threshold value for determining the state under residence are extracted as a residence cell list; and a residence position calculation part 13 for calculating a residence position for each residence cell list on the basis of a specified cell coordinate correspondence table or cell delay amount coordinate correspondence table and the residence period expectation values of the residence cell lists or the number of signals.

Description

  The present invention relates to a stay position estimation apparatus and a stay position estimation method for estimating a stay position of a terminal user staying in a cell based on position data obtained by cell-based or delay amount-based positioning.

  As a method for acquiring terminal user position data, for example, a technique for estimating the position of a terminal user in units of cells such as a cell-based positioning method, a technique for estimating the position of a terminal user in units of delay such as a PRACH-PD positioning method, In addition, a technique for estimating the position of a terminal user based on a GPS (Global Positioning System) signal (see Patent Document 1) is known, and cell-based positioning and delay-based positioning have relatively low accuracy, The latter GPS positioning is known to have a relatively high accuracy. Among these, in cell-based positioning, for example, a technique is known in which a cell in which a terminal user is located (hereinafter referred to as “location cell”) is determined, and position data of the terminal user is obtained from cell representative coordinates related to the cell. In the delay amount-based positioning, for example, a technique for acquiring terminal user position data from a cell where the terminal user is located and a delay amount between the terminal and the base station apparatus is known.

JP 2011-171876 A

  However, in an actual wireless network, there is a possibility that a “flapping phenomenon” in which the cell where the terminal user is located frequently switches due to the unstable reception environment of radio waves. If this phenomenon occurs, it appears that the location cell identifier and the associated cell representative coordinates are frequently changed even though the actual terminal has not moved from the same point. Therefore, it has been a problem to suppress the influence of the above phenomenon.

  On the other hand, in the conventional technology related to cell-based or delay-based positioning, a moving terminal user and a staying terminal user are treated as being the same without being distinguished from each other. However, it is difficult to obtain effective statistical information on tourism and traffic such as person trip statistics unless the terminal user's position is estimated by discriminating whether the terminal is moving or staying.

  Therefore, the present invention appropriately determines whether the mobile terminal is staying or staying based on the position data obtained by the cell-based or delay-based positioning, and estimates the average staying position of the terminal user who is staying. For the purpose.

  The stay position estimation apparatus according to the present invention provides terminal identification information, positioning time information, location cell identification information of the terminal, and the location of the terminal and the location cell obtained by cell-based or delay-based positioning for the target terminal. Processed data for each record of location cell data including the amount of delay between base stations, with the expected residence time in the location cell and the expected arrival time at the location cell calculated based on a predetermined rule. And a time-series list for each terminal representing a cell transition of the location cell of the terminal by extracting a record for each terminal based on terminal identification information from the processed data and rearranging based on the expected arrival time value To obtain a threshold for determining that the sum of expected residence time values is staying, dividing the time-series list into staying cell list candidates between neighboring cells based on the neighbor relationship information between the cells. A group extraction unit that extracts the staying cell list candidates as the staying cell list, a predetermined cell coordinate correspondence table or cell delay amount coordinate correspondence table for each extracted staying cell list, and the staying cell list A dwell position calculation unit that calculates a dwell position based on the expected dwell period value or the number of signals.

  In the stay position estimation device having such a configuration, the preprocessing unit obtains the terminal identification information, the positioning time information, the location cell identification information of the terminal obtained by the cell-based or delay amount-based positioning for the target terminal, and For each record of the location cell data including the amount of delay between the terminal and the base station of the location cell, the expected residence time in the location cell and the expected arrival time at the location cell are calculated based on a predetermined rule. To give machining data. Next, the group extraction unit extracts a record for each terminal based on the terminal identification information from the processed data, and rearranges it based on the expected arrival time value. Obtained, divide the time series list into adjacent cell list candidates between adjacent cells based on the adjacent relationship information between cells, and the remaining cell list candidate whose sum of expected retention period values is equal to or greater than the threshold for determining that the cells are staying As a staying cell list. Then, for each extracted staying cell list, the staying position calculation unit is based on a predetermined cell coordinate correspondence table or cell delay amount coordinate correspondence table and a staying period expected value or number of signals related to the staying cell list. The residence position is calculated.

  Thus, when extracting a staying cell list, a group extraction part divides | segments the time series list showing a location cell transition into the staying cell list candidates of adjacent cells based on the adjacent relationship information of cells. This makes it possible to identify a plurality of location cell data observed due to the flapping phenomenon or the like. Further, the group extraction unit extracts a staying cell list candidate as a staying cell list by extracting a staying cell list candidate whose sum of expected staying periods is equal to or more than a threshold for determining staying among the staying cell list candidates. / Appropriately discriminating the movement and appropriately extracting only the staying cell list candidates concerning the staying. As described above, it is possible to estimate the average staying position of the terminal user who is staying while appropriately determining whether it is moving or staying, while suppressing the influence of the flapping phenomenon or the like.

  The group extraction unit further includes at least one of a condition regarding the maximum value of the number of located cells included in a single staying cell list and a condition regarding the maximum value of the number of hops between the located cells. Extraction may be performed. In this case, for example, even if the terminal repeats positioning very frequently, it is possible to prevent inconvenience that the entire time-series list representing the location cell transition is extracted as one staying cell list.

  In addition, the preprocessing unit arranges the records of the location cell data in the time series of the positioning time, and when the (i + 1) th and (i-1) th records exist as expected residence period values for the i-th record, A value half the time difference between the positioning time of the (i + 1) th record and the positioning time of the (i-1) th record is calculated, and the (i + 1) th record exists and the (i-1) th record is If it does not exist, a value half the time difference between the positioning time of the (i + 1) th record and the positioning time of the ith record is calculated, and the (i-1) th record exists and the (i + 1) th record is If it does not exist, a value half the time difference between the positioning time of the i-th record and the positioning time of the (i−1) -th record may be calculated. Although details will be described later with reference to FIG. 8, the expected retention period value can be easily and appropriately calculated by the above method.

  In addition, the pre-processing unit arranges the records of the location cell data in the time series of the positioning time, and when the (i-1) th record exists, the positioning time of the i-th record and the (i-1) -th record The time at the apportioning point on the time series with the positioning time of is calculated as an expected arrival time for the i-th record, and if the (i-1) th record does not exist, the positioning time of the i-th record is i It may be the expected arrival time for the second record. In this way, the expected arrival time value can be calculated easily and appropriately. Note that the “prorated point on the time series” is a midpoint when the positioning time of the i-th record and the positioning time of the (i−1) -th record are arranged on the time axis. Alternatively, the points may be prorated by a predetermined ratio (for example, 6: 4).

  In addition, the stay position calculation unit calculates a weighted average of coordinates representing the representative position of the cell or a position corresponding to the delay amount using a weight based on the expected value of the stay period or the number of signals, and obtains the obtained weighted average for each stay It is good also as a residence position regarding a cell list. In this way, the staying position regarding each staying cell list can be appropriately calculated by a simple method.

  Note that the “adjacent relationship information between cells” may be determined in advance or may be derived as follows. That is, the adjacent relationship information between cells includes two cells related to adjacent records whose difference in positioning time of adjacent records is equal to or less than a predetermined value in a plurality of records in the time series list for each terminal as adjacent cells. It may be derived by judging. In this way, adjacent relationship information between cells can be derived by a simple method. In the derivation of the adjacent relationship information between the cells, the derivation based on the long-term time-series list for a larger number of users can be more accurately derived.

  The invention related to the stay position estimation device described above can also be regarded as an invention related to the stay position estimation method, and can be described as follows.

  The stay position estimation method according to the present invention is a stay position estimation method executed by a stay position estimation device, and is obtained by terminal-based terminal identification information obtained by cell-based or delay-based positioning for a target terminal, Expected residence time in the location cell calculated based on a predetermined rule for each record of positioning time information, location cell identification information, location cell data including the amount of delay between the terminal and the base station of the location cell And a pre-processing step for obtaining processed data by assigning an expected arrival time value to the location cell, and extracting records for each terminal based on the terminal identification information from the processed data and rearranging them based on the expected arrival time value Thus, a time series list for each terminal representing the cell transition of the location of the terminal is obtained, and the time series list is obtained based on the neighboring relationship information between the cells. A group extraction step of extracting as a staying cell list a staying cell list candidate that is equal to or greater than a threshold value for determining that the sum of the expected staying period values is staying, and predetermined for each extracted staying cell list A stay position calculation step of calculating a stay position based on the cell coordinate correspondence table or the cell delay amount coordinate correspondence table and the stay period expected value or the number of signals related to the stay cell list.

  According to the present invention, based on position data obtained by cell-based or cell delay amount-based positioning, it is possible to appropriately determine whether the terminal is moving or staying and estimate the average staying position of the terminal user who is staying. can do.

It is a functional block block diagram of the stay position estimation apparatus which concerns on embodiment of this invention. It is a figure which shows the hardware structural example of a stay position estimation apparatus. It is a flowchart which shows a stay position estimation process. It is a flowchart which shows the subroutine of a group extraction process. (A) is a figure for demonstrating location cell data, (b) is a figure for demonstrating process data. (A) is a figure for demonstrating an adjacent cell correspondence table, (b) is a figure for demonstrating group data. (A) is a diagram for explaining the cell coordinate correspondence table, (b) is a diagram for explaining the cell delay amount coordinate correspondence table, and (c) is a diagram for explaining the stay position data. is there. It is a figure for demonstrating calculation of the feature-value (expansion period expected value in a location cell), (a) is a figure in case the (i + 1) th and (i-1) th record exists, (b) (I + 1) th record exists but (i-1) th record does not exist, (c) shows (i-1) th record but (i + 1) th record exists The figure when not doing is shown, respectively. It is a figure which shows an example of the undirected graph expression of an adjacent cell corresponding table. It is a figure for demonstrating derivation | leading-out of an adjacent cell corresponding table.

  Embodiments according to the present invention will be described below with reference to the drawings.

[Configuration of residence position estimation device]
As shown in FIG. 1, the stay position estimation apparatus 10 according to the present embodiment includes a preprocessing unit 11, a group extraction unit 12, and a stay position calculation unit 13, and includes location cell data, an adjacent cell correspondence table, and cell coordinates to be described later. A correspondence table is input from the external system 20. The external system 20 can be configured by any one or more systems that store and manage the location cell data, the adjacent cell correspondence table, and the cell coordinate correspondence table.

  The pre-processing unit 11 calculates the “residence period expected value in the location cell” (hereinafter referred to as “below”) calculated for each record of the location cell data obtained by the cell-based or delay amount-based positioning for the target terminal. This is a part for executing pre-processing (step S1 in FIG. 3) for obtaining processed data by giving "feature value") and "expected time of arrival at the location cell" (hereinafter referred to as "arrival time"). is there. Specific examples of “location cell data” input to the preprocessing unit 11 and “processing data” output from the preprocessing unit 11 will be described later together with the description of the stay position estimation processing.

  The group extraction unit 12 uses the processing data described later and a predetermined adjacent cell correspondence table to perform group extraction processing for identifying a plurality of location cell identifiers observed for a terminal staying at a certain point (see FIG. 3 is a part for executing step S2). Specific examples of the “neighbor cell correspondence table” input to the group extraction unit 12 and the “group data” output from the group extraction unit 12 will be described later together with the description of the stay position estimation process. Note that the derived table may be used as the adjacent cell correspondence table in addition to a predetermined table, and the derivation method will be described later with reference to FIG.

  The stay position calculation unit 13 uses the group data described later and a predetermined cell coordinate correspondence table or cell delay amount coordinate correspondence table to represent the representative coordinates associated with the location cell identifier where the terminal staying at a certain point is located. In consideration of the staying period in each cell, the process of estimating the place where the terminal will stay (step S3 in FIG. 3) is executed. Specific examples of the “cell coordinate correspondence table” input to the stay position calculation unit 13 and the “stay position data” output from the stay position calculation unit 13 will be described later together with the description of the stay position estimation process.

  FIG. 2 shows an example of the hardware configuration of the stay position estimation apparatus 10. As the hardware configuration, the staying position estimation device 10 can be applied to a general information processing device (server, various types of computers such as a stationary type or a portable type). For example, as illustrated in FIG. A RAM 10B, a ROM 10C, an input device 10D such as a keyboard or mouse as input devices, a communication device 10E for communicating with an external device, an auxiliary storage device 10F, and an output device 10G such as a display or printer as output devices. With. The functions of the preprocessing unit 11, the group extraction unit 12, and the stay position calculation unit 13 of the stay position estimation device 10 described above cause the RAM 10B and the like to read a predetermined program and operate each device of FIG. 2 under the control of the CPU 10A. This is realized.

[Residence position estimation process]
Hereinafter, the process (stay position estimation process) according to the stay position estimation method of the present invention will be described. As shown in FIG. 3, the stay position estimation process includes a preprocess executed by the preprocessing unit 11 (step S1: preprocessing step) and a group extraction process executed by the group extracting unit 12 (step S2: group extraction step). ) And a stay position calculation process (step S3: stay position calculation step) executed by the stay position calculation unit 13. Hereinafter, the processing content of each step will be described.

(Step S1: Pre-processing step)
As shown in FIG. 1, the preprocessing unit 11 inputs the location cell data (FIG. 5A) from the external system 20, executes preprocessing described below, and processes data (FIG. 5B) as an execution result. ) Is output to the group extraction unit 12.

  The location cell data input to the preprocessing unit 11 is obtained by cell-based or delay-based positioning for the target terminal, and as shown in FIG. 5A, the terminal identifier, positioning time, location It consists of multiple records including cell identifier and delay amount. Here, the terminal identifier is a number for identifying each terminal, the positioning time is the time at which the location cell is located, the location cell identifier is a number for identifying the location cell, and The amount of delay is the amount of delay between the terminal and the base station of the location cell.

  The preprocessing unit 11 assigns a feature value w and an arrival time a calculated based on the following formula to each record of the location cell data, and obtains processed data.

Specifically, the preprocessing unit 11 inputs the location cell data from the external system 20, extracts a record relating to a certain terminal k from the location cell data using the terminal identifier as a key, and extracts the extracted record from the positioning time. Sort by t. Here, i is an index value corresponding to the rearrangement order. Further, the preprocessing unit 11 calculates the feature amount w _i by the following equation (1) and the arrival time a _i by the following equation (2) for the i-th record after the rearrangement, and the obtained feature amount Assign w _i and arrival time a _i to the i-th record. In the following formulas (1) and (2), the positioning time of the i-th record is t _i , the positioning time of the (i−1) -th record is t _i−1 , and the positioning time of the (i + 1) -th record is Let t _{i + 1} .

The preprocessing unit 11 performs the calculation of the feature value w _i and the arrival time a _i as described above and the assignment to the record for all records related to a certain terminal k. Furthermore, the preprocessing unit 11 obtains processed data by repeatedly executing it for all terminals and outputs the processed data to the group extraction unit 12.

  As shown in FIG. 5B, each record of processed data output from the preprocessing unit 11 includes a feature amount (retention in the location cell) in addition to the above-described terminal identifier, positioning time, location cell identifier, and delay amount. Expected period) and arrival time (expected arrival time to the cell).

Incidentally, when supplemented for formula (1), as shown in FIG. 8 (a), the midpoint between the (i-1) th record of the positioning time t _i-1 and i-th record of the positioning time t _i is P, when the i-th record of the positioning time _{t i} (i + 1) -th record midpoint between positioning time _{t i + 1} of the is Q, there is (i + 1) th and (i-1) th record The feature value w _i for the i-th record in this case is the time interval between the point P and the point Q, and the residence time (estimated that the terminal has stayed in the location cell corresponding to the location cell identifier of the i-th record) That is, it corresponds to the expected residence time).

However, when the (i + 1) th record exists but the (i-1) th record does not exist, as shown in FIG. 8B, the feature quantity w _i for the i-th record is i-th. and record the time interval between the positioning time t _i and the point Q of. On the other hand, when the (i-1) th record exists but the (i + 1) th record does not exist, as shown in FIG. 8C, the feature quantity w _i for the i-th record is the above point. and the time interval between the positioning time t _i P and i-th record.

Note that the method shown in the equation (1) and FIGS. 8A to 8C is an example of a method for calculating the feature value w _i , and is not limited to this method. Similarly, equation (2) is an example of a method for calculating arrival time a _i , and is not limited to this method.

(Step S2: Group extraction step)
As shown in FIG. 1, the group extraction unit 12 inputs the adjacent cell correspondence table (FIG. 6A) from the external system 20, and the processing data described above from the preprocessing unit 11, respectively. And the group data (FIG. 6B) is output to the stay position calculation unit 13 as an execution result.

  As shown in FIG. 6A, the neighbor cell correspondence table includes a plurality of records including a cell identifier that is a number for identifying a cell and a neighbor cell identifier that is a number for identifying a neighbor cell. ing.

さらに、グループ抽出部１２に入力される隣接セル対応表を無向グラフとして表現する。ここで、無向グラフの各ノードはセル識別子ｃを、各ノードをつなぐリンクは隣接セル対応表で定義されたノード間の隣接関係を、それぞれ表す。図９は、隣接セル対応表の無向グラフ表現の一例を示す。 The group extraction process includes three steps S21 to S23 shown in FIG. Hereinafter, the group extraction process will be described. Here, a new expression format of data is introduced. Here, C ^k , L ^k , W ^k , and A ^k are lists along the time series of the location cell identifier, delay amount, feature amount, and arrival time included in the processed data. As an example, assume that these lists are defined as follows: A numerical value in parentheses attached to the upper right of c represents a number for distinguishing cell identifiers.

Furthermore, the adjacent cell correspondence table input to the group extraction unit 12 is expressed as an undirected graph. Here, each node of the undirected graph represents a cell identifier c, and a link connecting each node represents an adjacent relationship between nodes defined in the adjacent cell correspondence table. FIG. 9 shows an example of an undirected graph representation of the adjacent cell correspondence table.

(Step S21)
In step S21 of FIG. 4, the group extraction unit 12 divides the location cell identifier list C ^k into partial lists based on the adjacent cell correspondence table. The divided partial list is referred to as “retained cell list C ^k _j ”. Here, in order to create the staying cell list C ^k _j (that is, to divide the location cell identifier list C ^k ), the group extraction unit 12 uses the adjacency relationship between successive elements (nodes) of the list C ^k ( As long as (link) is chained, these consecutive elements are stored in the same staying cell list C ^k _j , and when the chain ends, the subsequent elements are stored in a new staying cell list C ^k _{j + 1.} Apply repeatedly. This process is performed until the last element of the list C ^k is processed.

When the above-described processing is executed when the adjacent cell correspondence table of FIG. 9 and the above list C ^k are given, the following staying cell list C ^k _j is obtained.

  In the present embodiment, when creating a staying cell list in step S21, a restriction on the maximum number of nodes or the maximum link length included in one staying cell list is added. Hereinafter, how the staying cell list is created when restrictions on the maximum number of nodes and the maximum link length are added will be described with specific examples.

First, the operation when the maximum number of nodes is restricted will be described. When the maximum number of nodes is set to n, the group extraction unit 12 refers to the elements of the staying cell list (location cell identifiers) in order from the beginning, and when the number of different location cell identifiers exceeds n, the group extraction unit 12 Subdivide. According to this rule, when the adjacent cell correspondence table of FIG. 9 and the above list C ^k are given and the maximum node number n is 2, the following staying cell list is obtained.

Next, the operation when the maximum link length is restricted is described. When the maximum link length is set to n, the group extraction unit 12 refers to the elements of the staying cell list (location cell identifier) in order from the beginning, and when the number of hops between the location cell identifiers exceeds n, the staying cell list Subdivide. According to this rule, when the adjacent cell correspondence table of FIG. 9 and the above list C ^k are given, and the maximum link length n is 1, the following staying cell list is obtained.

  Since the restriction on the maximum number of nodes and the restriction on the maximum link length are similar, it is actually desirable to apply only one of the restrictions. When both constraints are applied, the behavior changes depending on which constraint is applied first. Therefore, it is determined in consideration of this point which constraint is applied first.

  As described above, when the restriction on the maximum number of nodes or the maximum link length included in one staying cell list is added, for example, for a terminal that repeats positioning very frequently, the entire time series list indicating the location cell transition Can be prevented from being extracted as one staying cell list. However, adding the restriction on the maximum number of nodes or the maximum link length as described above is not an essential requirement of the present invention.

(Step S22)
In step S _< b> 22, the group extraction unit 12 extracts a staying cell list C ^k _j whose staying period is equal to or greater than a predetermined threshold value for determining that the terminal is staying. In this process, feature amounts corresponding to each element of the stay cell list C ^k _j (same index j) are acquired from the feature amount list W ^k, and the total sum of the obtained feature amounts is set as a stay period p ^k _j .

By this step S22, the staying cell list (that is, the staying cell list determined to be staying) C ^k _j whose staying period is equal to or more than the above threshold, the delay amount list L ^k _j corresponding to the staying cell list, and the feature amount A list W ^k _j , a residence period p ^k _j, and an arrival time list A ^k _j are obtained.

(Step S23)
In step S23, the group extraction unit 12 processes the data obtained in step S22 to create the following group data. As shown in FIG. 6B, the group data includes a terminal identifier u ^k , a staying cell list C ^k _j obtained in step S22, a delay amount list L ^k _j corresponding to the staying cell list, and a feature amount. In addition to the list W ^k _j , the stay period p ^k _j and the arrival time list A ^k _j , the arrival time a ^k _j to the stay point and the departure time d ^k _j from the stay point are included.

Specifically, the group extraction unit 12 sets the value of the first element of the arrival time list A ^k _j as “arrival time a ^k _j ”, and adds the residence period p ^k _j to this arrival time a ^k _j to “departure” Time d ^k _j ”is obtained, and the group data shown in FIG. 6B is created and output to the stay position calculation unit 13.

(Step S3: Residence position calculation step)
As shown in FIG. 1, the stay position calculation unit 13 receives the cell coordinate correspondence table (FIG. 7A) or the cell delay amount coordinate correspondence table (FIG. 7B) from the external system 20 and the group extraction unit 12 described above. Each of the group data is input, a stay position calculation process described below is executed, and stay position data (FIG. 7C) is output as an execution result.

  As shown in FIG. 7A, the cell coordinate correspondence table includes a plurality of records including a cell identifier and coordinates representing the cell (cell representative coordinates (latitude) and cell representative coordinates (longitude)). . In addition, as shown in FIG. 7B, the cell delay amount coordinate correspondence table includes a cell identifier, a delay amount, and representative coordinates (that is, coordinates (latitude and longitude) of a position corresponding to the cell identifier and the delay amount). Consists of multiple records including

The stay position calculation process is a process of calculating stay position coordinates for each record of the group data. Specifically, for example, when the cell coordinate correspondence table is used, the stay position calculation unit 13 uses the cell identifier c that is an element of the stay cell list C ^k _j in the record having the group data as a key, from the cell coordinate correspondence table. Cell representative coordinates (latitude) and cell representative coordinates (longitude) are acquired and stored in a list format (X ^k _j , Y ^k _j ). On the other hand, when the cell delay amount coordinate correspondence table is used, the stay position calculation unit 13 uses the cell identifier c as an element of the stay cell list C ^k _j in the record having the group data as a key, and the cell delay amount coordinate correspondence table. The representative coordinates (latitude) and the representative coordinates (longitude) are acquired from and stored in a list format (X ^k _j , Y ^k _j ).

そして、滞留位置算出部１３は、滞留位置座標ｘ^＊ _ｊ，ｙ^＊ _ｊと、端末識別子ｕ^ｋ、到着時刻ａ^ｋ _ｊ、出発時刻ｄ^ｋ _ｊ、滞留期間ｐ^ｋ _ｊとを１つのレコードにまとめる。これにより、図７（ｃ）に示す滞留位置データの１つのレコードが作成される。 Thereafter, in common between the case of using the cell coordinate correspondence table and the case of using the cell delay amount coordinate correspondence table, the staying position calculation unit 13 uses the following formulas (3) and (4) to calculate the feature quantity list W ^k _j . A weighted average of X ^k _j and Y ^k _j is calculated as a weight, and the obtained coordinates (x ^* _j , y ^* _j ) are set as the staying position coordinates.

Then, the stay position calculation unit 13 combines the stay position coordinates x ^* _j , y ^* _j , the terminal identifier u ^k , the arrival time a ^k _j , the departure time d ^k _j , and the stay period p ^k _j into one record. . Thereby, one record of the staying position data shown in FIG. 7C is created.

  Residence position data for each stay location is created by executing the above-described stay position coordinate calculation for each record of the group data.

In the present embodiment described above, when the staying cell list is extracted by the group extraction process, the staying cell list C ^k _j is created by dividing the location cell identifier list C ^k based on the adjacent cell correspondence table. This makes it possible to identify a plurality of location cell data observed due to the flapping phenomenon or the like. In addition, among the staying cell list candidates, the staying cell list candidates that have a sum of expected staying period values equal to or greater than the threshold for determining staying are extracted as staying cell lists, so that the terminal staying / moving appropriately And only the staying cell list candidates related to staying are appropriately extracted. As described above, it is possible to estimate the average staying position of the terminal user in the staying state while appropriately determining whether the user is moving or staying, while suppressing the influence of the flapping phenomenon or the like.

In addition, as described above, by adding restrictions on the maximum number of nodes or the maximum link length included in one staying cell list, for example, for a terminal that repeats positioning very frequently, the entire list C ^{k of} located cell identifiers Can be prevented from being extracted as one staying cell list C ^k _j .

In step S3, an example of calculating a weighted average of X ^k _j and Y ^k _j using the feature quantity list W ^k _j as a weight, and using the obtained coordinates (x ^* _j , y ^* _j ) as the staying position coordinates. Although shown, this is an example of a method for calculating the stay position coordinates, and is not limited to this method. For example, when calculating the stay position coordinates by weighted average, the number of signals (that is, the number of signals related to the stay cell list C ^k _j whose stay period is equal to or greater than the threshold value) is not the feature amount (expected value of the stay period in the location cell). May be used. For example, the group extraction unit 12 counts the number of processed data records corresponding to the staying cell list C ^k _j whose staying period is equal to or greater than a threshold, and the number of signals is obtained as the number of records. What is necessary is just to pass to the stay position calculation part 13 as the number of signals regarding the cell list C ^k _j .

  In the above-described embodiment, an example of using a predetermined neighbor cell correspondence table has been described. However, the neighbor cell correspondence table may be derived as follows, in addition to using a predetermined neighbor cell correspondence table. That is, in a plurality of records in the time-series list for each terminal, two cells related to the adjacent records whose difference in positioning time between adjacent records is equal to or less than a predetermined reference value for determining the adjacent cells Are determined to be “adjacent cells”, and on the other hand, two cells related to the adjacent records whose positioning time difference between adjacent records is not less than or equal to the reference value are determined to be “not adjacent cells” Thus, an adjacent cell correspondence table may be derived.

For example, FIG. 10 shows a diagram in which a list C ^{k of the} cell identifiers of a certain terminal is represented on the time axis. Among the differences T _{1 to} T _{6 in} the positioning times of adjacent records, the differences T ₁ , T Assume that ₂ , T ₄ and T ₆ are equal to or less than the reference value. Hereinafter, in order to simplify the description, the cell with the cell identifier c ^(N) is referred to as “cell c ^(N) ”. For Figure 10, the cell ^{c (1), c (2} ) the difference between _T 1, _{T 4} corresponding to the movement between the since below the reference value, the cell ^{c (1), c (2} ) is determined as a neighboring cell To do. Similarly, cell ^{c (2), c (3)} the difference between _{T 2} corresponding to the movement between the since below the reference value, the cell ^{c (2), c (3)} is determined to be a neighboring cell, cell ^{c ( 4) Since} the difference T ₆ corresponding to the movement between c and c ⁽⁵⁾ is equal to or less than the reference value, the cells c ⁽⁴⁾ and c ⁽⁵⁾ are determined to be adjacent cells. On the other hand, since the difference T ₃ corresponding to the movement between the cells c ⁽³⁾ and c ⁽¹⁾ is not less than or equal to the reference value, the cells c ⁽³⁾ and c ⁽¹⁾ are determined not to be adjacent cells, and the cell c ^{( 2) Since} the difference T ₅ corresponding to the movement between c and ⁽⁴⁾ is not less than the reference value, it is determined that the cells c ⁽²⁾ and c ⁽⁴⁾ are not adjacent cells. An adjacent cell correspondence table may be derived from the information obtained as described above. If conflicting determination results are obtained, a larger number of determination results may be adopted.

  In this way, the neighbor cell correspondence table can be derived by a simple method. However, in the derivation of the adjacent cell correspondence table, the derivation based on the long-term time series list for a larger number of users can be more accurately derived.

DESCRIPTION OF SYMBOLS 10 ... Residence position estimation apparatus, 10A ... CPU, 10B ... RAM, 10C ... ROM, 10D ... Input device, 10E ... Communication device, 10F ... Auxiliary storage device, 10G ... Output device, 11 ... Preprocessing unit, 12 ... Group extraction Part, 13 ... stay position calculation part, 20 ... external system.

Claims (7)

Location cell including terminal identification information of the terminal, positioning time information, location cell identification information, and delay amount between the terminal and the base station of the location cell, obtained by the cell-based or delay amount-based positioning for the target terminal For each record of data, a pre-processing unit that obtains processing data by assigning an expected residence time value to the location cell and an expected arrival time value to the location cell calculated based on a predetermined rule;
By extracting records for each terminal based on the terminal identification information from the processed data and rearranging based on the expected arrival time value, a time series list for each terminal representing the location cell transition of the terminal is obtained, and between cells Based on the adjacency information, the time-series list is divided into the remaining cell list candidates between adjacent cells, and the remaining cell list candidate whose sum of the expected retention period values is equal to or greater than a threshold value for determining that it is staying A group extraction unit that extracts as
For each extracted staying cell list, a staying position for calculating a staying position based on a predetermined cell coordinate correspondence table or cell delay amount coordinate correspondence table and a staying period expected value or the number of signals related to the staying cell list. A calculation unit;
A stay position estimation apparatus comprising: The group extraction unit
The extraction of a staying cell list is further performed based on at least one of a condition relating to a maximum value of the number of located cells included in a single staying cell list and a condition relating to the maximum value of the number of hops between located cells. The residence position estimation apparatus described in 1. The pre-processing unit is
The records of the location cell data are arranged in the time series of the positioning time, and as the stay period expected value for the i-th record,
When the (i + 1) th and (i-1) th records exist, a value half the time difference between the positioning time of the (i + 1) th record and the positioning time of the (i-1) th record is calculated.
If the (i + 1) th record exists and the (i-1) th record does not exist, a value half the time difference between the positioning time of the (i + 1) th record and the positioning time of the ith record is calculated.
When the (i-1) th record exists and the (i + 1) th record does not exist, the half value of the time difference between the positioning time of the ith record and the positioning time of the (i-1) th record is calculated. The residence position estimation apparatus according to claim 1 or 2. The pre-processing unit is
Arrange the cell data records in chronological order of positioning time,
When the (i-1) -th record exists, the arrival time for the i-th record is determined by using the time distribution point between the positioning time of the i-th record and the positioning time of the (i-1) -th record. Calculate as the expected time,
(I-1) If the i-th record does not exist, the positioning time of the i-th record is set as the expected arrival time for the i-th record.
The stay position estimation apparatus according to any one of claims 1 to 3. The stay position calculation unit includes:
A weighted average of coordinates representing a representative position of a cell or a position corresponding to a delay amount is calculated using a weight based on the expected value of the stay period or the number of signals, and the obtained weighted average is used as a stay position for each staying cell list. The residence position estimation apparatus according to any one of claims 1 to 4. The adjacency information between the cells is
In a plurality of records in the time-series list for each terminal, derived by determining two cells related to the adjacent records whose difference in positioning time of adjacent records is a predetermined value or less as adjacent cells, The stay position estimation apparatus according to any one of claims 1 to 5. A stay position estimation method executed by a stay position estimation device,
Location cell including terminal identification information of the terminal, positioning time information, location cell identification information, and delay amount between the terminal and the base station of the location cell, obtained by the cell-based or delay amount-based positioning for the target terminal For each record of data, a pre-processing step of obtaining processed data by assigning a residence time expected value to the location cell calculated based on a predetermined rule and an arrival time expectation value to the location cell;
By extracting records for each terminal based on the terminal identification information from the processed data and rearranging based on the expected arrival time value, a time series list for each terminal representing the location cell transition of the terminal is obtained, and between cells Based on the adjacency information, the time-series list is divided into the remaining cell list candidates between adjacent cells, and the remaining cell list candidate whose sum of the expected retention period values is equal to or greater than a threshold value for determining that it is staying A group extraction step to extract as
For each extracted staying cell list, a staying position for calculating a staying position based on a predetermined cell coordinate correspondence table or cell delay amount coordinate correspondence table and a staying period expected value or the number of signals related to the staying cell list. A calculation step;
A residence position estimation method comprising:

Images