JP6329915B2 - Positioning system - Google Patents

Description

  The present invention relates to a positioning system.

  Outdoor positioning technology is very effective in terms of cost and positioning range thanks to the Global Positioning System (GPS). However, indoor positioning technology is very expensive because it requires a lot of infrastructure to be installed in places where GPS signals do not reach. Pedestrian location estimation (PDR) by dead reckoning is one method of indoor positioning technology. The PDR system uses a mobile sensor such as an accelerometer, a gyroscope, a magnetometer, or a barometer to estimate a traveling direction and a moving speed, and can estimate a relative position from an initial position based on the estimation result. The PDR system can be used together with a fixed indoor positioning system, thereby reducing the cost and extending the positioning range.

  Data collected by sensors such as accelerometers and gyroscopes will be different when the position of the sensor is in a different part of the body of the mobile terminal holder. In order to estimate the movement amount (movement speed and direction) of the owner of the mobile terminal, the possession position of the sensor is important information. Known methods use such information to correct the calculation of the amount of movement of the mobile terminal holder.

  For example, Patent Document 1 determines the magnitude of a swing motion of a mobile terminal, and when the magnitude of the swing motion is greater than or equal to a predetermined value, the mobile terminal performs the swing motion. And determining the portable position of the portable terminal in the pedestrian's body using at least one of the detected acceleration and angular velocity, and according to the determined portable position of the portable terminal It is disclosed to estimate the stride of (see summary).

  Patent Document 2 discloses that the positioning device calculates from the measurement data a holding state estimation unit that specifies how the device is held by the moving body, and a calculation according to the holding state that the holding state estimation unit specifies. Disclosed is a movement amount estimation unit that performs processing and calculates a movement vector of the moving body from the measurement data (see summary).

Special table 2013-533481 International Publication No. 2011/93447

  However, in estimating the amount of movement of the owner of the mobile terminal, not only the position of the sensor is important, but also the type of movement operation of the owner of the mobile terminal is extremely important. For example, the motions of walking, climbing stairs, and descending stairs show similar acceleration data, but the stride for walking and the stride for climbing up and down the stairs are completely different. Therefore, in order to correct the calculation for estimating the movement amount of the owner of the mobile terminal, both the possessing position of the sensor and the operation content of the owner are essential information.

  For example, using the technique disclosed in Patent Document 2, the possession position of the sensor device can be estimated, and the movement amount of the owner of the mobile terminal can be calculated using the information. However, if the owner's movement operation is unknown, the movement amount of the owner of the mobile terminal can be estimated incorrectly. For example, while walking and stair descent indicate similar acceleration sensor data, the walking stride is longer than the stair descent. Further, stair down includes movement in the vertical axis, which can be known only when the movement operation of the owner of the mobile terminal is given.

  Therefore, a technology that can appropriately estimate the position of the owner of the mobile terminal using data from the sensor of the mobile terminal is desired.

  A representative example of the present invention is a positioning system for determining the position of the owner of a mobile terminal, the measurement data measured by the mobile terminal and collected from the mobile terminal, and the measurement data by the mobile terminal The possession position detection model indicating the relationship between the feature amount in the mobile terminal and the possessed position of the mobile terminal, and corresponding to each of a plurality of possessed positions, and the relationship between the feature amount in the measurement data by the mobile terminal and the operation of the owner A possession position detection unit that determines a possession position of the mobile terminal based on a parameter table that stores a plurality of motion recognition models and a plurality of parameter sets, the possession position detection model, and the collected measurement data A model selection unit that selects a motion recognition model corresponding to the possessed position determined by the possessed position detection unit, and the selected motion recognition Based on Dell and the collected measurement data, the operation of the owner of the mobile terminal is determined, the operation recognition unit, the possession position determined by the possession position detection unit, and the operation determined by the operation recognition unit An adjustment unit that selects a parameter set from the parameter table, a movement amount calculation unit that calculates a movement amount of the holder using the selected parameter set, and the calculated movement amount And a position determination unit that determines the position of the holder based on

  According to one aspect of the present invention, it is possible to appropriately estimate the position of the mobile terminal owner using measurement data obtained by the sensor of the mobile terminal.

An example of composition of an indoor positioning system concerning Example 1 is shown typically. 1 shows functional blocks of an indoor positioning system according to a first embodiment. The measurement data concerning Example 1 are shown typically. The possession position detection model which concerns on Example 1 is shown typically. An example of an action recognition model concerning Example 1 is shown typically. An example of an action recognition model concerning Example 1 is shown typically. An example of an action recognition model concerning Example 1 is shown typically. An example of an action recognition model concerning Example 1 is shown typically. The parameter table concerning Example 1 is shown typically. 2 is a flowchart of possession position detection processing according to the first embodiment. 3 is a flowchart of model selection processing according to the first embodiment. 3 shows a flowchart of motion recognition processing according to the first embodiment. 3 shows a flowchart of an adjustment process according to the first embodiment. 3 is a flowchart of a movement amount calculation process according to the first embodiment. The flowchart of the position determination process of the holder of the mobile terminal which concerns on Example 1 is shown. The functional block diagram of the indoor positioning system which concerns on Example 2 is shown. FIG. 6 is a schematic diagram illustrating data collection according to the second embodiment. 9 shows a flowchart of a data collection unit according to a second embodiment. 9 is a flowchart of the possession position detection learning unit according to the second embodiment. The functional block diagram of the indoor positioning system 1 which concerns on Example 3 is shown. 9 shows a flowchart of a feedback system unit according to a third embodiment. The functional block diagram of the indoor positioning system which concerns on Example 4 is shown. 9 shows a flowchart of a data collection unit according to a fourth embodiment. 10 shows a flowchart of a motion recognition learning unit according to a fourth embodiment. The functional block diagram of the indoor positioning system which concerns on Example 5 is shown. 10 is a flowchart of a feedback system unit according to the fifth embodiment. The functional block diagram of the indoor positioning system which concerns on Example 6 is shown. It is typically shown how the correction part which concerns on Example 6 corrects possession position detection. It is typically shown how the correction part which concerns on Example 6 corrects action recognition. 20 shows an example of an image displayed on an input / output device of a server according to Embodiment 7. The other example of the application of the indoor positioning system which concerns on Example 7 is shown. The other example of the application of the indoor positioning system which concerns on Example 7 is shown.

  Embodiments of the present invention will be described below with reference to the accompanying drawings. It should be noted that this embodiment is merely an example for realizing the present invention, and does not limit the technical scope of the present invention. In each figure, the same reference numerals are given to common configurations.

  In the following, an indoor positioning system for estimating the position of the owner of a mobile terminal is disclosed. The indoor positioning system estimates the position of the owner of the mobile terminal based on the data measured by the sensor in the mobile terminal.

  The indoor positioning system estimates the possessed position of the mobile terminal by the owner by referring to the possessed position detection model from the measurement data of the mobile terminal. The indoor positioning system selects a motion recognition model corresponding to the estimated possessed position, and estimates the motion of the owner from the measurement data with reference to the selected motion recognition model.

  The indoor positioning system selects a parameter set corresponding to the estimated possessed position and operation from the parameter table. The indoor positioning system estimates the moving speed and traveling direction of the owner from the measurement data obtained by the mobile terminal using the selected parameters. The indoor positioning system calculates the amount of movement of the owner from the estimated moving speed and direction of travel.

  The indoor positioning system of the present disclosure uses a motion recognition model corresponding to the estimated possession position of the mobile terminal to more accurately estimate the owner's motion, and from the estimated possession position and motion, the movement speed and The amount of movement of the owner determined from the traveling direction can be estimated more accurately.

  A first embodiment will be described with reference to FIGS. When the position of the mobile terminal holder is estimated based on measurement data measured by a measurement sensor such as an accelerometer, a gyroscope, a magnetometer, or a barometer included in the mobile terminal, the adjustment program Adjust the parameters to calculate the amount of movement of the person regularly. The movement amount includes an element in the movement direction. The parameter adjustment improves the accuracy of estimation of the movement amount of the owner of the mobile terminal. As a result, the position of the owner of the moving mobile terminal can be accurately estimated.

  FIG. 1 schematically illustrates a configuration example of an indoor positioning system according to the first embodiment. The indoor positioning system includes a server 100 and a mobile terminal 200. The mobile terminal 200 includes a measurement sensor group 230. The measurement sensor group 230 includes, for example, an accelerometer that detects acceleration due to movement of the owner of the mobile terminal 200, a gyroscope that detects angular velocity when the owner of the mobile terminal 200 changes its orientation, and a magnetic field around the mobile terminal 200. And a barometer that detects the atmospheric pressure around the mobile terminal 200.

  The mobile terminal 200 stores measurement data 241 measured by the measurement sensor group 230. The server 100 collects measurement data 241 measured by the measurement sensor group 230 of the mobile terminal 200 and calculates the amount of movement of the owner based on the collected measurement data 241.

  The mobile terminal 200 includes a processor 210, a memory 220, a measurement sensor group 230, an auxiliary storage 240, a communication interface 250, and an input / output device 260. These components are connected to each other by a bus. A smartphone is an example of the mobile terminal 200. The mobile terminal 200 is not limited to a smartphone as long as measurement data of movement of the owner can be measured by the measurement sensor group 230.

  The processor 210 refers to the memory 220 and executes various arithmetic processes. The memory 220 stores a measurement data acquisition program 221. The measurement data acquisition program 221 stores the measurement data collected by the measurement sensor group 230 in association with the measurement time. Measurement data associated with the measurement time is stored in the auxiliary storage 240 as measurement data 241. The measurement data includes, for example, acceleration due to movement of the owner of the mobile terminal 200, angular velocity due to change of the direction of the owner of the mobile terminal 200, magnetic field around the mobile terminal 200, atmospheric pressure around the mobile terminal 200, and the like.

  The measurement sensor group 230 can measure, for example, acceleration, angular velocity, magnetic field, atmospheric pressure, and the like of the mobile terminal 200. A typical accelerometer measures accelerations of three axes orthogonal to each other, specifically, the vertical axis, the horizontal axis, and the vertical axis of the frame of the mobile terminal 200. A typical gyroscope measures the angular velocities of three orthogonal axes, specifically, the vertical axis, the horizontal axis, and the vertical axis of the frame of the mobile terminal 200. A typical magnetometer measures magnetic fields on three orthogonal axes, specifically, the vertical axis, the horizontal axis, and the vertical axis of the frame of the mobile terminal 200. The barometer measures the atmospheric pressure around the mobile terminal 200.

  The auxiliary storage 240 stores measurement data 241. The auxiliary storage 240 is, for example, a portable atmospheric pressure storage medium. Details of the measurement data 241 will be described later with reference to FIG. The communication interface 250 is an interface that connects the mobile terminal 200 to the network 150. The input / output device 260 is a device for receiving input from the user and presenting information to the user, and includes, for example, an input key, a touch panel, a display, and the like.

  The server 100 includes a processor 110, a memory 120, an auxiliary storage 130, a communication interface 140, and an input / output device 160. These components are connected to each other by a bus. The processor 110 refers to the memory 120 and executes various arithmetic processes. The memory 120 stores a possessed position detection program 121, a model selection program 122, an action recognition program 123, an adjustment program 124, a movement amount calculation program 125, and a position determination program 126.

  Based on the measurement data 132, the possession position detection program 121 detects where the mobile terminal 200 is possessed by the owner of the mobile terminal 200. The measurement data 132 corresponds to the measurement data 241 collected in the mobile terminal 200 and is stored in the auxiliary storage 130 in the server 100. A collection program (not shown) of the server 100 collects measurement data 241 from the mobile terminal 200 and stores it as measurement data 132. Processing by the possession position detection program 121 will be described later with reference to FIG.

  The model selection program 122 selects an appropriate model for the motion recognition program 123 based on the possessed position detected by the possessed position detection program 121. Many motion recognition models are stored in the auxiliary memory 130, and each motion recognition model corresponds to the position of the mobile terminal 200 possessed by the owner. These possession positions are, for example, on the owner's hand, in the owner's belt pouch, in the owner's side pocket, in the owner's front pocket, in the owner's front pocket, and the like. The processing by the model selection program 122 will be described later with reference to FIG.

  The motion recognition program 123 determines the motion of the owner of the mobile terminal 200 based on the motion recognition model selected by the measurement data 132 and the model selection program 122. The movement is, for example, walking, running, standing still, going up stairs, going down stairs, opening a door, moving by an elevator, and the like. The processing by the motion recognition program 123 will be described later with reference to FIG.

  The adjustment program 124 changes parameters for estimating the stride, the number of steps, the traveling direction, the number of floors, and the like based on the possessed position and the information of the motion recognition. The parameters to be adjusted include, for example, a parameter used for step count counting, a parameter used for step length calculation, a parameter used for determination of a traveling direction through a door, and a parameter used for rank determination. The processing of the adjustment program 124 will be described later with reference to FIG.

  The movement amount calculation program 125 calculates the movement speed and traveling direction of the holder of the mobile terminal 200. The amount of movement can be calculated from the moving speed and traveling direction of the holder of the mobile terminal 200.

  The moving speed is a moving distance per unit time, and the moving distance can be calculated from, for example, the number of steps and the step length of the owner of the mobile terminal 200. The number of steps is determined by, for example, the number of cycles of the acceleration signal. The stride can be estimated from, for example, the number of steps per unit time and the reference stride of the owner of the mobile terminal 200.

  The traveling direction can be calculated based on the direction of the mobile terminal 200 with respect to the earth or a change in direction. For example, the magnetometer indicates a direction with reference to the east, west, north, and south of the mobile terminal 200, and the change in direction can be calculated from the measurement data of the gyroscope. The movement amount calculation program 125 will be described in detail later with reference to FIG.

  The position determination program 126 determines the position of the owner of the mobile terminal 200. The position of the owner of the mobile terminal 200 can be traced from the initial position, the traveling direction, and the movement amount calculated by the movement amount calculation program 125. The position determination program 126 can also determine the position of the owner of the mobile terminal 200 based on a fixed access point such as a WIFI access point or a beacon access point.

  The auxiliary memory 130 stores a motion recognition model group 131, and each motion recognition model is associated with each position on the body of the owner of the mobile terminal 200. The auxiliary storage 130 further stores measurement data 132, a possession position detection model 133, and a parameter table 134.

  Many motion recognition models can be defined, for example, a motion recognition model on the hand, a motion recognition model in the belt pouch, a motion recognition model in the side pocket, a motion recognition model in the front pocket, and the like. Each motion recognition model stores a feature amount for recognizing the motion of the owner of the mobile terminal 200 based on the measurement data 132. Examples of the motion recognition model included in the motion recognition model group 131 will be described later with reference to FIGS.

  The measurement data 132 is sensor data and corresponds to the measurement data 241 in the mobile terminal 200. Specifically, the server 100 periodically transmits a measurement data acquisition request to the mobile terminal 200 via the network 150. When receiving the measurement data acquisition request, the mobile terminal 200 transmits the collected measurement data 241 to the server 100 via the network 150. When the measurement data 241 is received from the mobile terminal 200, the server 100 stores the received measurement data 241 in the auxiliary storage 130. An example of the measurement data 132 will be described later with reference to FIG.

  The method by which the server 100 collects the measurement data 241 of the mobile terminal 200 is not limited to the above method. The server 100 may directly read the measurement data 241 from the portable storage medium in which the measurement data 241 is stored.

  The possession position detection model 133 is stored in order to determine the possession position of the mobile terminal 200 in the owner's body. An example of the possession position detection model 133 will be described later with reference to FIG. The parameter table 134 stores parameters for calculating the amount of movement of the owner of the mobile terminal 200. An example of the parameter table 134 will be described later with reference to FIG.

  The communication interface 140 is an interface for connecting the server 100 to the network 150. The input / output device 160 is a device that receives input from the user and presents information to the user, and includes, for example, a mouse, input keys, a touch panel, a display, and the like.

  The indoor positioning system is a computer that executes the possession position detection program 121, the model selection program 122, the motion recognition program 123, the adjustment program 124, the movement amount calculation program 125, and the position determination program 126, and is not limited to the server 100. . For example, when the mobile terminal 200 executes the possessed position detection program 121, the model selection program 122, the motion recognition program 123, the adjustment program 124, the movement amount calculation program 125, and the position determination program 126, the mobile terminal 200 It has the same function and operates as an indoor positioning system.

  FIG. 2 illustrates functional blocks of the indoor positioning system according to the first embodiment. The indoor positioning system 1 includes a storage unit 2 and a control unit 3. The storage unit 2 stores all data, specifically, a motion recognition model group 131, measurement data 132, a possessed position detection model 133, and a parameter table 134. The control unit 3 includes all processing units, specifically, a possessed position detection unit 1210, a model selection unit 1220, an operation recognition unit 1230, an adjustment unit 1240, a movement amount calculation unit 1250, and a position determination unit 1260.

  The possessed position detection unit 1210 is realized by the processor 110 executing the possessed position detection program 121. The possession position detection unit 1210 reads the measurement data 132, compares the feature amount calculated from the measurement data 132 with the possession position detection model 133, and determines the possession position of the mobile terminal 200 in the body of the owner of the mobile terminal 200. judge. The measurement data 132 is collected from the mobile terminal 200 by a collection unit realized by the processor 110 that executes a collection program (not shown).

  The possession position detection unit 1210 extracts a feature amount from the measurement data 132 by performing statistical mathematical operations on the measurement data 132. The possession position detection unit 1210 compares the calculated feature quantity with the feature quantity in the possession position detection model 133 to determine the possession position of the mobile terminal 200 in the owner's body. The processing of the possession position detection unit 1210 will be described in detail later with reference to FIG.

  The model selection unit 1220 selects an appropriate motion recognition model based on information from the possession position detection unit 1210. The model selection unit 1220 is realized by the processor 110 executing the model selection program 122. The process of the model selection unit 1220 will be described later with reference to FIG.

  The action recognition unit 1230 determines the action of the owner of the mobile terminal 200 based on the measurement data 132. The motion recognition unit 1230 is realized by the processor 110 executing the motion recognition program 123. The motion recognition unit 1230 extracts feature amounts from the measurement data 132 by performing statistical and arithmetic calculations on the measurement data. The motion recognition unit 1230 compares the calculated feature amount with the selected motion recognition model to determine the motion of the owner of the mobile terminal 200. The process of the action recognition unit 1230 will be described later with reference to FIG.

  The adjustment unit 1240 changes the parameter used for calculating the movement amount according to the parameter table 134. The adjustment unit 1240 is realized by the processor 110 executing the adjustment program 124. The adjustment unit 1240 receives information from the possession position detection unit 1210 and the motion recognition unit 1230, selects a parameter from the parameter table 134, and replaces the parameter used in the previous movement amount calculation. The processing of the adjustment unit 1240 will be described later with reference to FIG.

  The movement amount calculation unit 1250 calculates the movement speed and traveling direction of the owner of the mobile terminal 200. The movement amount calculation unit 1250 is realized by the processor 110 executing the movement amount calculation program 125.

  The movement amount calculation unit 1250 calculates the movement amount every predetermined time (for example, 3 seconds). A sampling window is defined by the predetermined time. The movement amount calculation unit 1250 counts the number of steps according to the period of the acceleration data within the sampling window, and further estimates the step length of each step based on the acceleration data.

  The acceleration data is one of the data types in the measurement data 132. Based on the number of steps, the stride, and the sampling window time (predetermined time), the movement amount calculation unit 1250 can estimate the movement distance (movement speed) per unit time of the owner of the mobile terminal 200.

  The movement amount calculation unit 1250 can calculate the change in the traveling direction from the traveling direction in the previous movement amount calculation by integrating the angular velocity data with time. Angular velocity data is one of the data types in the measurement data 132. The traveling direction can also be determined from magnetic data. The processing of the movement amount calculation unit 1250 will be described later with reference to FIG.

  The position determination unit 1260 determines the position of the holder of the mobile terminal 200. The position determination unit 1260 is realized by the processor 110 executing the position determination program 126. The position determination unit 1260 can estimate the position of the owner of the mobile terminal 200 from the initial position and the initial traveling direction and the route calculated by the movement amount calculation unit 1250.

  However, if there is a reset point whose exact position is known, the position determination uses the correct position determined from the reset point instead of the estimated position. The reset point is a fixed access point such as, for example, a WIFI access point or a beacon access point, and the position of the owner of the mobile terminal 200 can be determined more accurately based on the signal strength from the access point. The processing of the position determination unit 1260 will be described later with reference to FIG.

  FIG. 3 schematically shows the measurement data 132 according to the first embodiment. Measurement data 132 may include any type of sensor data. Examples of sensors include an accelerometer, a gyroscope, a magnetometer, and a barometer. However, the measurement sensor group 230 can include sensors other than those described above as long as it can be mounted on the mobile terminal 200.

  The measurement data 132 includes a sensor type column 301, a time column 302, an X-axis value column 303, a Y-axis value column 304, and a Z-axis value column 305. The sensor type column 301 indicates the type of sensor. Each sensor is associated with a unique identifier. For example, an acceleration sensor is indicated by an “A” identifier, a gyroscope is indicated by a “Y” identifier, a magnetic sensor is indicated by a “T” identifier, and a barometer is indicated by a “B” identifier. The sensor is not limited to the above sensor. The other sensor data is, for example, received signal strength from a WIFI access point or a Bluetooth access point (Bluetooth is a registered trademark, the same applies hereinafter).

  The time when the sensor data is measured is stored in the time column 302. Data measured by the sensor is stored in an X-axis value field 303, a Y-axis value field 304, and a Z-axis value field 305.

  The X axis, the Y axis, and the Z axis are orthogonal to each other. For example, when the sensor type is an accelerometer, the lateral acceleration of the mobile terminal 200 is stored in the X axis value field 303, and the vertical direction of the mobile terminal 200 is The acceleration in the vertical direction with respect to the mobile terminal 200 is stored in the Z-axis value column 305. The same applies to other sensor data types. The barometer data indicates values common to the X axis, the Y axis, and the Z axis.

  FIG. 4 schematically illustrates the possession position detection model 133 according to the first embodiment. The possession position detection model 133 is referred to in order to determine the possession position of the mobile terminal 200. The possession position detection model 133 stores data for determining the possession position in a specific one motion, for example, a walking motion. The possessed position detection model 133 includes a position column 401 and a detected feature amount column, specifically, a minimum acceleration column 402, a maximum acceleration column 403, and an acceleration standard deviation column 404.

  The position column 401 stores the possessed position of the mobile terminal 200, that is, the place where the mobile terminal 200 is arranged. The holder of the mobile terminal 200 can hold the mobile terminal 200 on a hand, put it in a belt pouch, put it in a front pocket, put it in a side pocket, or put it in a bag. The registered positions are not limited to these, and may be any positions where the mobile terminal 200 can be placed when the owner moves.

The minimum acceleration column 402 stores the minimum value of the magnitude of acceleration in the sampling window. As shown in the following equation 1, the magnitude of acceleration is calculated as a normalized value of the sum of squares of acceleration values in all axial directions at the same measurement time.

  As is the magnitude of acceleration. Ax is an acceleration value in the X-axis direction, Ay is an acceleration value in the Y-axis direction, and Az is an acceleration value in the Z-axis direction.

  The maximum acceleration column 403 stores the maximum acceleration magnitude in the sampling window. The acceleration standard deviation column 404 stores the standard deviation of the magnitude of acceleration in the sampling window.

  All the feature values are stored for determining the possession position of the mobile terminal 200. In this example, the model is stored as a decision tree model, and the possessed position of the mobile terminal 200 is classified using the feature amount of acceleration data as an input. Each cell of the feature amount columns 402 to 404 stores a feature amount range or a feature amount threshold value. The possessed position detection unit 1210 can calculate the feature amount of the acceleration data from the measurement data 132, compare the feature amount of the acceleration data with the possessed position detection model 133, and determine the possessed position of the mobile terminal 200.

  For example, the record in which the cell of the minimum acceleration column 402 includes the minimum acceleration of the acceleration data is selected. In the selected record, the record in which the cell of the maximum acceleration column 403 includes the maximum acceleration of the acceleration data is selected. Further, in the selected record, the record in which the cell of the acceleration standard deviation column 404 includes the acceleration data standard deviation is selected.

  Note that the feature amount for detecting the possessed position of the mobile terminal 200 is only an example. Other types of feature quantities for detecting the possessed position of the mobile terminal 200 may be used. The possession position detection model 133 is not limited to a decision tree, and any learning model can be used. The decision tree or the determination table is an example of a model that can distinguish the possessing position of the mobile terminal 200. Another example of the possession position detection model 133 is a support vector machine model, which stores a support vector instead of a range of each feature value.

  5A to 5D schematically show an example of the motion recognition model according to the first embodiment stored for each possessed position. Depending on the number of possessed positions that can be distinguished and detected by possessed position detection, a plurality of models for motion recognition are prepared. Each motion recognition model corresponds to one possessed position.

  5A to 5D show four motion recognition models 131A to 131D, which are stored for each possessed position of the mobile terminal 200. FIG. Specifically, the motion recognition model 131A on the hand, the motion recognition model 131B in the belt pouch, the motion recognition model 131C in the side pocket, and the motion recognition model 131D in the front pocket. For example, different models are stored as different databases or different files with different file names, or stored in the same database with different identifiers.

  The motion recognition models 131A to 131D have the same structure. An example of the structure includes an action column 501 and a feature amount column for recognizing the action, for example, an acceleration standard deviation column 502 and an atmospheric pressure change column 503. The operation column 501 shows the operation of the owner of the mobile terminal 200. The movement operation of the owner of the mobile terminal 200 is, for example, walking, running, going up stairs, going down stairs, going up by an elevator, going down by an elevator, opening a door, and the like.

  The acceleration standard deviation column 502 indicates the standard deviation of the magnitude of acceleration in the sampling window. The atmospheric pressure change column 503 stores the atmospheric pressure change amount in the sampling window.

  All feature quantities are stored for determining the operation of the owner of the mobile terminal 200. In this example, the model is stored as a decision tree model, and the movement of the owner is classified with the feature values of acceleration data and atmospheric pressure data as inputs. Each cell in the feature amount columns 502 and 503 stores a feature amount range or a feature amount threshold value.

  For example, in the atmospheric pressure change column 503, the atmospheric pressure change of walking, running, and stationary is defined as a range of −0.5 to +0.5. The change in atmospheric pressure due to stairs and elevators is defined as a range greater than 0.5 and less than or equal to 1. The change in atmospheric pressure due to stairs and elevators is defined as a range smaller than −0.5 and −1 or more. The value (range) in the atmospheric pressure change column 503 is, for example, common to the motion recognition models 131A to 131D.

  For the same motion, the standard deviation of acceleration varies depending on the motion recognition model model. When the mobile terminal 200 is put in a pants pocket, the mobile terminal 200 is closer to the owner's foot, and therefore, the larger movement of the mobile terminal 200 is recorded.

  As a result, for example, the walking acceleration standard deviation in the motion recognition model 131A on the hand is smaller than the walking acceleration standard deviation in the side pocket motion recognition model 131D. For example, the walking acceleration standard deviation in the motion recognition model 131A on the hand is defined as a range of 0.2 to 0.4, and the walking acceleration standard deviation in the side pocket motion recognition model 131D is from 0.3. It is defined as a range of 0.5.

  The motion recognition unit 1230 calculates the feature amount of the acceleration data and the atmospheric pressure data from the measurement data 132, compares the calculated feature amount with the motion recognition model corresponding to the possessed position of the mobile terminal 200, and the owner of the mobile terminal 200 Can be determined.

  FIG. 6 schematically illustrates the parameter table 134 according to the first embodiment. The parameter table 134 stores a parameter set used for calculating the movement amount of the owner of the mobile terminal 200. The parameter table 134 stores one parameter set for each combination of all possessing positions of the mobile terminal 200 and all operations of the owner. The movement amount calculation unit 1250 selects an estimated possession position and parameter set of motion from the parameter table 134, and calculates the movement amount using the selected parameter set.

  In the example of FIG. 6, the parameter table 134 stores two parameters for estimating the stride and one parameter for estimating the altitude change. The parameter table 134 includes a position column 601, an action column 602, a first step length parameter column 603, a second step length parameter column 604, and an altitude change column 605. The position column 601 indicates the possessed position of the mobile terminal 200, and the operation column 602 indicates the operation of the owner of the mobile terminal 200.

  The first parameter column 603 for stride and the second parameter column 604 for stride store parameters for calculating the stride. The stride is calculated using, for example, acceleration data in the sampling window and the stride basic value of the owner of the mobile terminal 200. The altitude change column 605 stores the altitude change amount for the first floor. Specifically, an up / down record by a staircase or an elevator indicates the altitude change amount for the first floor.

  FIG. 7 is a flowchart of the possession position detection process according to the first embodiment. The possession position detection process is executed by the processor 110 of the server 100 that operates as the possession position detection unit 1210. FIG. 7 illustrates an example of classifying the possessed position of the mobile terminal 200 using the measurement data 132.

  The possessed position detection unit 1210 may determine the possessed position of the mobile terminal 200 by another method. For example, the possession position detection unit 1210 detects the nearest possession position using a sensor outside the mobile terminal 200 attached to the body of the owner of the mobile terminal 200.

  The possession position detection unit 1210 reads the measurement data 132 and filters it into each type of data. For example, the data is classified into acceleration data, gyroscope data, magnetometer data, and barometer data (step 701).

  The possession position detection unit 1210 uses the measurement data 132 to determine whether the owner is walking. One determination method detects a period in acceleration data and determines whether the owner is walking from the period. The possession position detection unit 1210 can determine whether or not the acceleration data indicates a walking motion by using the autocorrelation method (step 702).

  If movement due to walking is not detected, the possession position detection unit 1210 reads other measurement data until movement due to walking can be detected (step 701).

  On the other hand, when walking is detected, the possession position detection unit 1210 compares the feature amount of the magnitude of acceleration calculated by the above equation 1 with the possession position detection model 133 (step 703). The possession position detection unit 1210 determines which possession position (for example, on the hand or in the side pocket) the measurement data 132 indicates, for example, according to the decision tree model as described above (step 704).

  Thus, the possession position detection process ends. When the possession position detection model 133 is based on an action other than walking, the possession position detection unit 1210 detects the action instead of walking. A plurality of possession position detection models 133 for different operations may be prepared.

  FIG. 8 is a flowchart illustrating model selection processing according to the first embodiment. The model selection process is executed by the processor 110 of the server 100 that operates as the model selection unit 1220. The model selection unit 1220 receives the possessed position determined by the possessed position detection unit 1210 (step 801).

  The model selection unit 1220 selects an action recognition model corresponding to the possessed position. For example, when the possession position detection unit 1210 determines that the mobile terminal 200 is on the hand of the owner, the model selection unit 1220 receives information from the possession position detection unit 1210 that the possession position is “hand”. To get. The model selection unit 1220 selects the motion recognition model 131A on the hand from the motion recognition model group 131, and loads the motion recognition model 131A on the hand from the auxiliary storage 130 into the memory 120.

  After a while, when the owner puts the mobile terminal 200 in the side pocket, the possession position detection unit 1210 detects the possession position in the side pocket in the corresponding sampling window. The model selection unit 1220 acquires information that the possessed position is “side pocket”. The model selection unit 1220 loads the motion recognition model 131D in the side pocket from the auxiliary storage 130 into the memory 120 (step 802). Thus, the model selection process ends.

  FIG. 9 is a flowchart of the motion recognition process according to the first embodiment. The motion recognition process is executed by the processor 110 of the server 100 that operates as the motion recognition unit 1230. The motion recognition unit 1230 acquires the motion recognition model loaded in the model selection process (step 901).

  Next, the motion recognition unit 1230 reads the measurement data 132 and classifies it into each type of data. For example, the measurement data is classified into acceleration data, angular velocity data, magnetic data, and atmospheric pressure data (step 902).

  The motion recognition unit 1230 calculates the feature amount of the measurement data 132, and compares the feature amount of the measurement data 132 with the selected motion recognition model (step 903). The action recognition unit 1230 determines which action of the owner of the mobile terminal 200 indicates the measurement data 132 according to, for example, the decision tree model as described above (step 904). Thus, the process ends.

  FIG. 10 is a flowchart of the adjustment process according to the first embodiment. The adjustment process is executed by the processor 110 of the server 100 that operates as the adjustment unit 1240.

  The adjustment unit 1240 acquires information about the possessed position and the owner's motion from the possessed position detecting unit 1210 and the motion recognizing unit 1230 (step 1001). Next, the adjustment unit 1240 determines whether the possessed position has changed from the state determined in the previous sampling window (step 1002).

  If the possessed position has changed from the previous state (step 1002: YES), the adjustment unit 1240 searches the parameter table 134 using the new possessed position and action combination as a key, and acquires the corresponding parameter set ( Step 1003). The adjustment unit 1240 changes the parameter with a new parameter set (step 1004). If the possessed position has not changed from the previous state (step 1002: NO), no parameter change is necessary. Thus, the process ends.

  FIG. 11 is a flowchart of the movement amount calculation process according to the first embodiment. The movement amount calculation process is executed by the processor 110 of the server 100 that operates as the movement amount calculation unit 1250.

  The movement amount calculation unit 1250 reads the measurement data 132 and classifies it into each type of data. For example, the movement amount calculation unit 1250 classifies the measurement data 132 into acceleration data, angular velocity data, magnetic data, and atmospheric pressure data (step 1101).

  Next, the movement amount calculation unit 1250 can estimate the movement speed of the owner of the mobile terminal 200 and the traveling direction or a change thereof based on the measurement data 132. For example, the movement amount calculation unit 1250 counts the number of steps in the sampling window from the period of the acceleration data in the measurement data 132.

  Further, the movement amount calculation unit 1250 can estimate the stride of each step based on the magnitude of acceleration at each step. Parameters for calculating the stride based on the possessed position of the mobile terminal 200 and the operation of the owner are selected in advance by the adjustment unit 1240 from the parameter table 134. The movement amount calculation unit 1250 can calculate the movement speed of the owner of the mobile terminal 200 from the stride and the number of steps (step 1102).

  Furthermore, the movement amount calculation unit 1250 can estimate the traveling direction or change in the traveling direction of the owner of the mobile terminal 200 from the magnetic data or the angular velocity data (step 1103). The movement amount calculation unit 1250 can calculate the movement route of the owner of the mobile terminal 200 by integrating the movement speed and the movement direction change with time (step 1104). Thus, the movement amount calculation process ends.

  FIG. 12 is a flowchart of the position determination process for the owner of the mobile terminal 200 according to the first embodiment. The position determination process is executed by the processor 110 of the server 100 that operates as the position determination unit 1260.

  The position determination unit 1260 determines whether the current position of the owner of the mobile terminal 200 can be estimated directly from the reset point (step 1201). The reset point is a position that is accurately known, for example, a fixed access point. For example, when the owner of the mobile terminal 200 approaches a known WIFI access point, the received signal strength of the mobile terminal 200 is within a predetermined range in which the position of the owner of the mobile terminal 200 is close to the WIFI access point. You can show that there is.

  In another example, when the mobile terminal 200 can receive an outdoor GPS signal, and when the mobile terminal 200 starts to move indoors, the initial position or reset point is the entrance of the building, The direction of travel is the direction to enter the building.

  When the position determination unit 1260 finds a reset point (S1201: YES), the position determination unit 1260 determines the position of the owner of the mobile terminal 200 based on the reset point (step 1202). The process ends as described above.

  When the position determination unit 1260 does not find the reset point (S1201: NO), the owner of the mobile terminal 200 based on the position determination unit 1260, the position determined in the previous sampling window, and the amount of movement from the previous position. Is calculated (step 1203). The movement amount is calculated by the movement amount calculation unit 1250 as described above. Thereafter, the position determination unit 1260 displays the new position of the owner of the mobile terminal 200 on the display device of the input / output device 160, or transmits data for display to the mobile terminal 200 (step 1204). The process ends here.

  As described above, according to the present embodiment, the position of the owner of the mobile terminal 200 can be appropriately estimated from the sensor data of the mobile terminal 200.

  A second embodiment will be described with reference to FIGS. In the first embodiment, a possession position detection model 133 is prepared in advance and stored in the storage unit 2. In this embodiment, an example of a method for generating the possession position detection model 133 will be described. In this embodiment, the possession position detection model 133 is generated by learning processing.

  FIG. 13 is a functional block diagram of the expanded indoor positioning system 1 according to the second embodiment. The components are the same as the functional blocks of the indoor positioning system of the first embodiment except that the learning processing unit 4 is added. The learning processing unit 4 is realized by, for example, the processor 110 executing a learning processing program.

  The learning processing unit 4 includes a data collection unit 1270 and a possession position detection learning unit 1280. The data collection unit 1270 collects measurement data at a plurality of possession positions of the mobile terminal. The data collection unit 1270 calculates the feature amount from the measurement data at each possessed position. The data collection unit 1270 associates the feature amount with the possessed position, and inputs them to the possessed position detection learning unit 1280. The flow of the data collection unit 1270 will be described later in detail with reference to FIG.

  The possession position detection learning unit 1280 generates the possession position detection model 133 from the feature amount at each possession position input from the data collection unit 1270. The flow of the possession position detection learning unit 1280 will be described later in detail with reference to FIG.

  FIG. 14 is a schematic diagram illustrating data collection according to the second embodiment. The data collection target 1410 is a person. Data is collected from many subjects. Explain target data collection. The smartphone is an example of the mobile terminal 200 and is attached to the target. In FIG. 14, the smartphone 1420 is on the hand of the subject 1410. Smartphone 1430 is in the front pocket of the subject shirt. Smartphone 1440 is in the target belt pouch. Smartphone 1430 is in the side pocket of subject 1410.

  As long as the body or clothes of the subject 1410 can hold a smartphone, the position and number of smartphones to be worn are not limited. The subject 1410 can have one or more smartphones.

  All smartphones are installed with software for collecting sensor data of the measurement sensor group 230. For example, accelerometers, gyroscopes, magnetometers, barometers, and sensor data such as WIFI or Bluetooth received signal strength are collected. These are examples of sensor data, and sensor data is not limited to these. Object 1410 is walking while the smartphone collects data.

  FIG. 15 shows a flowchart of the data collection unit 1270. As illustrated in FIG. 14, the smartphones 1420 to 1450 collect sensor data while the object 1410 is walking at different possession positions. The data collection unit 1270 collects measurement data (sensor data) collected by the smartphones 1420 to 1450 from the smartphones 1420 to 1450 and stores the data in the memory 120 for subsequent processing (step 1501).

  The data collection unit 1270 classifies all collected data into measurement data of actual possession positions of the smartphones 1420 to 1450 (step 1502). Next, the data collection unit 1270 divides the measurement data at each actual possessed position into segment data of the same size depending on the sampling window (step 1503).

  Finally, the data collection unit 1270 calculates the feature amount of each segment data in the measurement data at each possessed position. For example, a standard deviation of the magnitude of acceleration, a minimum value of the magnitude of acceleration, and a maximum value of the magnitude of acceleration are calculated. The magnitude of acceleration can be calculated according to Equation 1 shown in the first embodiment. The process ends as described above.

  FIG. 16 shows a flowchart of the possession position detection learning unit 1280. The possession position detection learning unit 1280 acquires the feature amount calculated by the data collection unit 1270 and associated with the actual possession position from the data collection unit 1270 (step 1601).

  The possession position detection learning unit 1280 calculates a possession position detection model 133 for classifying the possession position from the measurement data 132 based on the machine learning model (step 1602). Any type of classification model can be used as the learning model. For example, in addition to the decision tree described with reference to FIG. 4, a k-nearest neighbor method or a support vector machine can be used.

  After the possession position detection model 133 is calculated, the possession position detection learning unit 1280 stores it in the storage unit 2 (auxiliary storage 130) (step 1603). The process ends as described above. Instead of or in addition to walking, a possessed position detection model 133 for operations other than walking may be generated. In each operation, sensor data from the smartphones 1420 to 1450 is collected.

  According to the present embodiment, the possession position detection model 133 can be appropriately generated from the sensor data measured at the actual possession position.

  A third embodiment will be described with reference to FIGS. In the second embodiment, in order to detect the possession position of the mobile terminal 200, a lot of target data is collected and the possession position detection model 133 is calculated. In the third embodiment, the possession position detection model 133 is adjusted by the feedback system unit 5 to enable more accurate detection of the possession position in a specific target.

  FIG. 17 is a functional block diagram of the expanded indoor positioning system 1 according to the third embodiment. Components other than the feedback system unit 5 are the same as the functional blocks of the indoor positioning system according to the second embodiment. The feedback system unit 5 uses the possession position detection learning unit 1280 to update the possession position detection model 133 to a more accurate model of the mobile terminal 200.

  FIG. 18 shows a flowchart of the feedback system unit 5. The feedback system unit 5 requests the owner of the mobile terminal 200 to input information about the actual possession position, and receives feedback about the possession position from the owner (step 1801).

  For example, the feedback system unit 5 transmits an information input request to the mobile terminal 200 via the network 150, and the mobile terminal 200 displays a request for the owner in the input / output device 260. The owner inputs information on the actual possession position at the input / output device 260 of the mobile terminal 200, and the mobile terminal 200 transmits the information to the server 100 via the network 150.

  Next, the feedback system unit 5 compares the actual possessed position input by the owner of the mobile terminal 200 with the possessed position detected by the possessed position detection unit 1210 using the possessed position detection model 133. (Step 1802). If the possessed position is accurately detected by the possessed position detection unit 1210 (step 1802: YES), the process ends.

  If the possessed position is not accurately detected by the possessed position detection unit 1210 (step 1802: NO), the feedback system unit 5 adjusts the possessed position detection learning unit 1280 so that the possessed position detection model 133 is correctly updated. (Step 1803).

  For example, the feedback system unit 5 inputs information on the correct possession position from the owner to the possession position detection learning unit 1280. The possessed position detection learning unit 1280 receives, as inputs, feature amounts in a large number of sampling windows in the measurement data 132 and information on actual possessed positions from the owner of the mobile terminal 200, and records the possessed positions in the possessed position detection model 133. Update. The possession position detection learning unit 1280 also updates records at other possession positions as necessary.

By inputting feature amount measurement data of more sampling windows for more actual possessed positions as feedback, the possessed position detection learning unit 1280 can calculate the possessed position detection model 133 more accurately. The update of the possession position detection model 133 may be executed by other methods. For example, a penalty method may be used that penalizes the wrong classification.

  The indoor positioning system 1 may hold the possession position detection model 133 unique to the owner and adjust the possession position detection model 133 of the owner according to the feedback of the owner input from the feedback unit 5.

  As described above, according to the present embodiment, information on the actual possession position of the mobile terminal 200 is acquired from the possessor, and the possession position detection model 133 is made a more appropriate model based on the actual possession position and the measurement data 132. Can be updated.

  A fourth embodiment will be described with reference to FIGS. In the first embodiment, the motion recognition model group 131 is prepared in advance and stored in the storage unit 2. In this embodiment, an example of a method for generating the motion recognition model group 131 will be described. In this embodiment, the motion recognition model group 131 is generated by learning processing.

  FIG. 19 is a functional block diagram of the expanded indoor positioning system 1 according to the fourth embodiment. The constituent elements are the same as the functional blocks of the indoor positioning system 1 of the first embodiment except that the learning processing unit 4 is added. The learning processing unit 4 is realized by, for example, the processor 110 executing a learning processing program.

  The learning processing unit 4 includes a data collection unit 1271 and an action recognition learning unit 1281. The data collection unit 1271 collects measurement data of a plurality of operations at each of a plurality of possessed positions of the mobile terminal. The data collection unit 1270 calculates a feature amount from measurement data of each of a plurality of operations at each possessed position. The data collection unit 1271 associates the feature amount with the possessed position and the motion and inputs them to the motion recognition learning unit 1281. The flow of the data collection unit 1271 will be described later in detail with reference to FIG.

  The motion recognition learning unit 1281 generates motion recognition models 131 </ b> A to 131 </ b> D at each possessed position from the feature amounts of the motion at each possessed position input from the data collection unit 1271. The flow of the motion recognition learning unit 1281 will be described later in detail with reference to FIG.

  FIG. 20 shows a flowchart of the data collection unit 1271. As in the second embodiment, the smartphones 1420 to 1450 are held at different possession positions of the target 1410. Unlike the second embodiment, the smartphones 1420 to 1450 collect sensor data in different operations of the target 1410.

  The data collection unit 1271 collects measurement data (sensor data) collected by the smartphones 1420 to 1450 from the smartphones 1420 to 1450, and stores the data in the memory 120 for subsequent processing (step 2001).

  The data collection unit 1271 classifies all collected data according to pairs of actual possession positions and actions. Specifically, the data collection unit 1271 classifies the measurement data of the actual possession positions of the smartphones 1420 to 1450, and further classifies the measurement data of each possession position into the measurement data of the actual operation (Step 2002). .

  Next, the data collection unit 1271 divides the measurement data of each actual operation at each actual possessed position into segment data of the same size depending on the sampling window (step 2003).

  Finally, the data collection unit 1271 calculates the feature amount of each segment data in the measurement data of each actual operation at each actual possessed position. For example, the standard deviation of the magnitude of acceleration and the change in atmospheric pressure are calculated. The magnitude of acceleration can be calculated according to Equation 1 shown in the first embodiment. The process ends as described above.

  FIG. 21 shows a flowchart of the motion recognition learning unit 1281. The motion recognition learning unit 1281 obtains the feature amount calculated by the data collection unit 1271 and associated with each actual motion at each actual possessed position from the data collection unit 1271 (step 2101).

  The motion recognition learning unit 1281 calculates a motion recognition model group 131 for classifying the possessed position from the measurement data 132 based on the machine learning model. For example, the motion recognition learning unit 1281 generates the motion recognition models 131A to 131D based on the machine learning model from the feature amounts of different motions measured by the smartphones 1420 to 1450, respectively (step 2102).

  Any type of classification model can be used as the learning model. For example, in addition to the decision tree described with reference to FIG. 4, a k-nearest neighbor method or a support vector machine can be used. After the motion recognition model group 131 is calculated, the motion recognition learning unit 1281 stores it in the storage unit 2 (auxiliary storage 130) (step 2103). The process ends as described above.

  According to the present embodiment, the motion recognition model group 131 can be appropriately generated from the sensor data measured during the actual motion at the actual possessed position.

  A fifth embodiment will be described with reference to FIGS. In Example 4, in order to estimate the operation of the mobile terminal 200, a lot of target data is collected, and the operation recognition model group 131 is calculated. In the fifth embodiment, the motion recognition model group 131 is adjusted by the feedback system unit 5 to enable more accurate detection of the possessed position in a specific target.

  FIG. 22 is a functional block diagram of the expanded indoor positioning system 1 according to the fifth embodiment. Components other than the feedback system unit 5 are the same as the functional blocks of the indoor positioning system 1 according to the fourth embodiment. The feedback system unit 5 adjusts the motion recognition learning unit 1281 to make the motion recognition model group 131 more accurate.

  FIG. 23 shows a flowchart of the feedback system unit 5. The feedback system unit 5 requests the owner of the mobile terminal 200 to input information on the actual possession position and operation, and receives feedback on the possession position and operation from the owner (step 2301).

  For example, the feedback system unit 5 transmits an information input request to the mobile terminal 200 via the network 150, and the mobile terminal 200 displays a request for the owner in the input / output device 260. The owner inputs the actual possession position and operation information at the input / output device 260 of the mobile terminal 200, and the mobile terminal 200 transmits the information to the server 100 via the network 150.

  Next, the feedback system unit 5 includes the actual possessed position and the actual operation input by the owner of the mobile terminal 200, and the operation determined by the operation recognition unit 1230 using the operation recognition model group 131. Compare (step 2302). If the motion is accurately determined by the motion recognition unit 1230 (step 2302: YES), the processing ends.

  If the motion is not accurately determined by the motion recognition unit 1230 (step 2302: NO), the feedback system unit 5 adjusts the possession position detection learning unit 1280 so that the motion recognition model of the possession position is correctly updated. (Step 1803).

  For example, the feedback system unit 5 inputs the correct possession position and motion information from the owner to the motion recognition learning unit 1281. The motion recognition learning unit 1281 receives the feature values in a large number of sampling windows in the measurement data 132, the actual possession position from the owner of the mobile terminal 200, and information on the actual motion as inputs, and the possession in the motion recognition model group 131. Update the motion record of the motion recognition model of the position. The action recognition learning unit 1281 also updates records in other actions as necessary.

By inputting feature sampling data of more sampling windows for more actual possessed positions and actual motions as feedback, the possessed position detection learning unit 1280 can calculate the motion recognition model group 131 more accurately. Can do. The update of the motion recognition model group 131 may be executed by another method. For example, a penalty method may be used that penalizes the wrong classification.

  The indoor positioning system 1 holds a motion recognition model group 131 and a possession position detection model 133 specific to the owner, and according to the owner's feedback input from the feedback unit 5, the motion recognition model group 131 and The possession position detection model 133 may be adjusted.

  As described above, according to the present embodiment, information on the actual possession position of the mobile terminal 200 is acquired from the owner, and the motion recognition model group 131 is converted into a more appropriate model based on the actual possession position and the measurement data 132. Can be updated.

  Example 6 will be described with reference to FIGS. In the first embodiment, the possession position detection unit 1210 and the motion recognition unit 1230 may classify the possession position and the operation by mistake. The correction unit 1241 increases the accuracy of the indoor positioning system 1.

  FIG. 24 is a functional block diagram of the expanded indoor positioning system 1 according to the sixth embodiment. Except for the correction unit 1241, the components are the same as those in the indoor positioning system 1 according to the first embodiment. The correction unit 1241 is realized, for example, when the processor 110 of the server 100 operates according to a program.

  The correction unit 1241 corrects an erroneous classification by the possessed position detection unit 1210 and the motion recognition unit 1230. For example, a Kalman filter or a Markov model can be applied to the correction unit 1241. Correction of elements in a sequence using a Kalman filter or a Markov model is a well-known technique and will not be described. The correction unit 1241 is not limited to these, and it is sufficient that the erroneous classification of the possession position detection unit 1210 and the action recognition unit 1230 can be corrected.

  FIG. 25 schematically shows how the correction unit 1241 corrects the possession position detection. The possessed position classification result sequence 2501 indicates a sequence before correction. Each cell indicates the possessed position determined in the sampling window. The sequence 2501 includes an erroneous determination result “side pocket”. The possessing positions before and after the “side pocket” are determined to be “hands”.

  The correction unit 1241 can correct an incorrect sequence 2501 to a correct sequence 2502 using a Markov filter or a Kalman filter. Specifically, the correction unit 1241 uses a Markov filter or a Kalman filter to correct the “side pocket” to “hand” from the possession position before and after the “side pocket”.

  FIG. 26 schematically shows how the correction unit 1241 corrects motion recognition. An operation classification result sequence 2601 indicates a sequence before correction. Each cell represents an action determined in the sampling window. The sequence 2601 includes an erroneous determination result “step up”. The operation before and after “step up” is determined to be “walking”.

  The correction unit 1241 can correct an incorrect sequence 2601 to a correct sequence 2602 using a Markov filter or a Kalman filter. Specifically, the correction unit 1241 corrects “step up” to “walk” from the operation before and after “step up” using a Markov filter or a Kalman filter.

  The correction unit 1241 passes the correction result of the possession position or movement classification result sequence to the adjustment unit 1240. In this example, the adjustment unit 1240 passes the parameter set to the movement amount calculation unit 1250 for each sequence including a plurality of sampling windows, and the movement amount calculation unit 1250 calculates the movement amount for each sequence. A parameter set is selected for each sampling window.

  When the adjustment unit 1240 receives the corrected possession position sequence or operation sequence from the correction unit 1241, the adjustment unit 1240 selects a corresponding parameter set from the parameter table 134 for each sampling window in the sequence. As a result, the adjustment unit 1240 can select a parameter set corresponding to the correct possession position for the movement amount calculation unit 1250.

  A seventh embodiment will be described with reference to FIGS. FIG. 27 illustrates an application example of the indoor positioning system according to the seventh embodiment. The purpose of the application is to monitor workers in fields where GPS signals do not reach. For example, it can be used in a building, tunnel, or near. According to the present embodiment, the position of the owner of the mobile terminal 200 can be presented to the user in the field where the GPS signal does not reach.

  FIG. 27 shows an example image 2700 displayed on the input / output device 160 of the server 100, for example. An image 2700 shows information on each floor being monitored. In the example of FIG. 27, information on the first floor 2701 and information on the second floor 2702 are presented. On each floor, a map and the position of worker 2703 are displayed. The image shows walls and other obstacles 2704 as rectangles.

  The map information is stored in the auxiliary storage 130 of the server 100, for example. The server 100 displays the operation of the worker 2703 in the image 2700. Walking is shown as “W”, stationary as “S”, and running as “R”. In addition, server 100 plots the worker's recent path in image 2700. The position, movement and path of the worker 2703 are estimated as described in the above embodiment.

  FIG. 28 illustrates another example of the application of the indoor positioning system according to the seventh embodiment. The purpose of the application is smartphone user navigation. The smartphone 2800 guides the user to a desired destination based on the map information and the user position information.

  The auxiliary storage 240 of the smartphone 2800 stores the installed navigation program. The processor 210 provides a navigation service to the user by operating according to the program. As described in the above embodiment, the server 100 estimates the current position of the user and transmits the estimation result to the smartphone 2800 together with the map information. The smartphone 2800 displays the user position together with the map information on a display device in the input / output device.

  FIG. 29 illustrates another example of the application of the indoor positioning system according to the seventh embodiment. The purpose of the application is navigation by the voice of the smartphone user 2910. The input / output device of the smartphone 2900 includes a speaker. The smartphone 2900 guides the user 2910 to a desired destination based on the user's current position and map information acquired from the server 100 by voice from a speaker.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Further, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

  Each of the above-described configurations, functions, processing units, and the like may be realized by hardware by designing a part or all of them, for example, by an integrated circuit. Each of the above-described configurations, functions, and the like may be realized by software by interpreting and executing a program that realizes each function by the processor. Information such as programs, tables, and files for realizing each function can be stored in a memory, a hard disk, a recording device such as an SSD (Solid State Drive), or a recording medium such as an IC card or an SD card.

  In addition, the control lines and information lines are those that are considered necessary for the explanation, and not all the control lines and information lines on the product are necessarily shown. In practice, it may be considered that almost all the components are connected to each other.

4 learning processing unit, 5 feedback system unit, 131 motion recognition model group, 132 measurement data, 133 possession position detection model, 134 parameter table, 200 mobile terminal, 1210 possession position detection unit, 1220 model selection unit, 1230 motion recognition unit, 1240 adjustment unit, 1241 correction unit, 1250 movement amount calculation unit, 1260 position determination unit, 1270, 1271 data collection unit, 1280 possession position detection learning unit, 1281 motion recognition learning unit

Claims (11)

A positioning system for determining the position of a mobile terminal holder,
Measurement data measured by the mobile terminal and collected from the mobile terminal;
A possession position detection model showing a relationship between a feature amount in measurement data by the mobile terminal and a possession position of the mobile terminal;
A plurality of motion recognition models corresponding to each of a plurality of possessed positions, and indicating a relationship between a feature amount in measurement data by the mobile terminal and the motion of the owner;
A parameter table that stores multiple parameter sets;
A possession position detection unit for determining a possession position of the mobile terminal based on the possession position detection model and the collected measurement data;
A model selection unit that selects a motion recognition model corresponding to the possessed position determined by the possessed position detection unit;
A motion recognition unit for determining a motion of the owner of the mobile terminal based on the selected motion recognition model and the collected measurement data;
An adjustment unit that selects a parameter set from the parameter table based on the possession position determined by the possession position detection unit and the operation determined by the operation recognition unit;
A movement amount calculation unit for calculating the movement amount of the holder using the selected parameter set;
A position determination unit that determines the position of the owner based on the calculated movement amount;
A learning processing unit;
A feedback system section;
A correction section, and
The feedback system unit receives an input of an actual possession position by the owner from the mobile terminal, and inputs information on the actual possession position to the learning processing unit,
The learning processing unit adjusts the possession position detection model based on the actual possession position, the collected measurement data, and the possession position detection model,
The correction unit determines whether the first operation in the operation sequence of the holder is correct or incorrect based on operations before and after the first operation, and determines that the first operation is an error. Positioning system that corrects the operation of the . The positioning system according to claim 1,
A learning processing unit for generating the possession position detection model;
The learning processing unit
Obtain measurement data from multiple mobile terminals at different possession positions,
Calculating the feature quantity of the measurement data at each of the different possessed positions;
Based on the feature amount of each of the different possession positions, determine the relationship between each of the different possession positions and the feature amount in the measurement data,
A positioning system that generates the possessed position detection model based on a result of the determination.   The positioning system according to claim 1,
  A learning processing unit for generating the plurality of motion recognition models;
  The learning processing unit
  Obtain measurement data during different operations from multiple mobile terminals in different possession positions,
  Calculating feature quantities of the measurement data in each of the different possession positions and the different pairs of different movements;
  Based on the feature amount of each of the different pairs, determine the relationship between each of the different pairs and the feature amount in the measurement data,
  A positioning system that generates the plurality of motion recognition models based on a result of the determination.   The positioning system according to claim 1,
  A learning processing unit and a feedback system unit;
  The feedback system unit receives an input of an actual possession position and an actual operation by the owner from the mobile terminal, and inputs information on the actual possession position and the actual operation to the learning processing unit,
  The positioning system, wherein the learning processing unit adjusts the motion recognition model of the possession position based on the actual possession position, the actual motion, the collected measurement data, and the motion recognition model of the actual possession position.   The positioning system according to claim 1,
  A correction part,
  The correction unit determines whether the first possession position in the sequence of possession positions of the mobile terminal is correct based on possession positions before and after the first possession position, and the first possession position is incorrect. A positioning system that corrects the first possessed position upon determination.   The positioning system according to claim 1,
  A positioning system that outputs the position of the owner determined by the position determination unit together with map information.   A positioning method by a positioning system for determining the position of the owner of a mobile terminal,
  The positioning system is
  Measurement data measured by the mobile terminal and collected from the mobile terminal;
  A possession position detection model showing a relationship between a feature amount in measurement data by the mobile terminal and a possession position of the mobile terminal;
  A plurality of motion recognition models corresponding to each of a plurality of possessed positions, and indicating a relationship between a feature amount in measurement data by the mobile terminal and the motion of the owner;
  A parameter table for storing a plurality of parameter sets;
  The positioning system is
  Based on the possession position detection model and the collected measurement data, determine the possession position of the mobile terminal,
  Select a motion recognition model corresponding to the determined possessed position,
  Based on the selected movement recognition model and the collected measurement data, determine the movement of the owner of the mobile terminal,
  Based on the determined possessed position and the determined operation, a parameter set is selected from the parameter table,
  Using the selected parameter set, calculate the amount of movement of the holder,
  Determining the position of the holder based on the calculated amount of movement;
  Receiving an input of an actual possession position by the owner from the mobile terminal;
  Based on the actual possession position, the collected measurement data and the possession position detection model, adjust the possession position detection model,
  The correctness of the first action in the sequence of actions of the holder is determined based on the actions before and after the first action, and if the first action is determined to be incorrect, the first action is corrected. The positioning method.   The positioning method according to claim 7,
  The positioning system is
  Obtain measurement data from multiple mobile terminals at different possession positions,
  Calculating the feature quantity of the measurement data at each of the different possessed positions;
  Based on the feature amount of each of the different possession positions, determine the relationship between each of the different possession positions and the feature amount in the measurement data,
  A positioning method further comprising: generating the possession position detection model based on the result of the determination.   The positioning method according to claim 7,
  Obtain measurement data during different operations from multiple mobile terminals in different possession positions,
  Calculating feature quantities of the measurement data in each of the different possession positions and the different pairs of different movements;
  Based on the feature amount of each of the different pairs, determine the relationship between each of the different pairs and the feature amount in the measurement data,
  A positioning method further comprising: generating the plurality of motion recognition models based on a result of the determination.   The positioning method according to claim 7,
  Receiving the actual possession position and the actual movement input by the owner from the mobile terminal;
  The positioning method further comprising adjusting the motion recognition model of the actual possession position based on the actual possession position, the actual motion, the collected measurement data, and the motion recognition model of the actual possession position.   The positioning method according to claim 7,
  When the correctness of the first possessed position in the possessed position sequence of the mobile terminal is determined based on the possessed positions before and after the first possessed position, and the first possessed position is determined to be incorrect, A positioning method further comprising correcting the possessing position of one.

Images