JP6271793B2 - Determination device, determination method, and determination program - Google Patents

Description

  The present invention relates to a determination device, a determination method, and a determination program.

  2. Description of the Related Art Conventionally, a car navigation technology (hereinafter also referred to as “guidance”) that guides a vehicle on which a user is boarding to a destination using a portable terminal device such as a smartphone is known. A terminal device that performs such guidance uses a satellite positioning system such as GPS (Global Positioning System) to identify the current location of the vehicle, and displays a map or guidance route screen and the identified current location in an overlapping manner. .

  On the other hand, the terminal device cannot display the current location in a place where it is difficult to receive a signal from a satellite such as in a tunnel. The same problem is not limited to GPS, but is common to all positioning using external signals (for example, radio waves from mobile phone (cellular) base stations, wireless LAN radio waves, etc.). Therefore, an autonomous positioning technique that estimates the current location of the vehicle using the acceleration measured by the accelerometer is conceivable. For example, a method has been proposed in which a device having an accelerometer is fixed in a vehicle in a predetermined posture, and a traveling state of the vehicle is determined from acceleration detected by the device.

Japanese Patent No. 4736866

  However, the terminal device such as a smartphone has a different installation posture in the vehicle depending on the type of the vehicle on which the user gets on, the usage status of the holder that holds the terminal device, etc. Judgment may not be possible.

  For example, since the terminal device has an accelerometer that detects axial acceleration with respect to the terminal device, the direction of the detected acceleration is based on the moving direction of the vehicle based on the installation posture of the terminal device. Convert to And a terminal device specifies the advancing direction and speed of a vehicle using the acceleration which changed the direction, and specifies the present location of a vehicle from the specified advancing direction and speed. However, when the installation posture is unknown or when the installation posture changes, the terminal device cannot change the direction of acceleration, and the determination accuracy of the running state deteriorates.

  The present application has been made in view of the above, and an object thereof is to provide a determination device, a determination method, and a determination program that can easily determine the traveling state regardless of the installation posture of the terminal device.

  The determination apparatus according to the present application includes a detection unit that detects acceleration, a setting unit that sets a reference direction based on the acceleration detected by the detection unit at a preset time interval, and a reference direction set by the setting unit An acquisition unit that acquires a feature amount based on an acceleration in a direction with reference to the position, and a determination unit that determines a moving state of the terminal device using the feature amount acquired by the acquisition unit. .

  According to one aspect of the embodiment, there is an effect that the traveling state of the vehicle can be determined regardless of the installation posture of the terminal device.

Drawing 1 is a figure for explaining an example of an operation effect which a terminal unit concerning an embodiment exhibits. FIG. 2 is a diagram illustrating an example of a functional configuration of the terminal device according to the embodiment. FIG. 3 is a diagram illustrating an example of information registered in the learning data database according to the embodiment. FIG. 4 is a diagram illustrating an example of information registered in the test data database according to the embodiment. FIG. 5 is a diagram illustrating an example of information registered in the model database according to the embodiment. FIG. 6 is a flowchart for explaining an example of a flow of guidance processing executed by the terminal device according to the embodiment. FIG. 7 is a flowchart for explaining an example of a flow of determination processing executed by the terminal device according to the embodiment. FIG. 8 is a flowchart for explaining an example of the flow of acquisition processing executed by the terminal device according to the embodiment. FIG. 9 is a flowchart illustrating an example of the flow of learning processing executed by the terminal device according to the embodiment. FIG. 10 is a diagram illustrating an example of a time until the stop determination is started.

  Hereinafter, a mode for carrying out a determination device, a determination method, and a determination program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the determination device, the determination method, and the determination program according to the present application are not limited by this embodiment. Further, in the following embodiments, the same parts and processes are denoted by the same reference numerals, and redundant description is omitted.

  In the following description, an example of car navigation that guides a vehicle on which a user has boarded to a destination will be described as processing executed by the determination device, but the embodiment is not limited thereto. For example, the determination device executes processing described below and guides the user to the destination even when the user is walking or using a transportation means other than a vehicle such as a train. May be executed.

[1. (Overview of moving state)
First, the concept of the movement mode determined by the terminal device 10 which is an example of a determination device will be described with reference to FIG. Drawing 1 is a figure for explaining an example of an operation effect which a terminal unit concerning an embodiment exhibits. For example, the terminal device 10 is a mobile terminal such as a smartphone, a tablet terminal or a PDA (Personal Digital Assistant), or a terminal device such as a notebook PC (Personal Computer), such as a mobile communication network or a wireless LAN (Local Area Network). The terminal device can communicate with an arbitrary server via the network N.

  Further, the terminal device 10 has a car navigation function for guiding the vehicle C10 on which the user has boarded to the destination. For example, when receiving an input of a destination from the user, the terminal device 10 acquires route information for guiding the user to the destination from a server or the like (not shown). For example, the route information includes a route to a destination that can be used by the vehicle C10, information on an expressway included in the route, traffic jam information on the route, information on a facility serving as a guide, information on a map displayed on the screen, Data such as images such as voice and maps output at the time of guidance are included.

  Further, the terminal device 10 has a positioning function for specifying the position of the terminal device 10 (hereinafter referred to as “current location”) at a predetermined time interval using a satellite positioning system such as GPS (Global Positioning System). . Then, the terminal device 10 displays an image such as a map included in the route information on a liquid crystal screen, electroluminescence, an LED (Light Emitting Diode) screen or the like (hereinafter simply referred to as “screen”) and specifies the image. The current location is displayed on the map each time. Further, the terminal device 10 displays a left turn or a right turn, a change in lane to be used, an estimated arrival time at the destination, or the like according to the specified current location, or is output by voice from a speaker of the terminal device 10 or the vehicle C10. To do.

  Here, in the satellite positioning system, signals transmitted from a plurality of satellites are received, and the current location of the terminal device 10 is specified using the received signals. For this reason, the terminal device 10 cannot identify the current location when it cannot properly receive a signal transmitted from a satellite, such as a place in a tunnel or a building group. Moreover, the application etc. which implement | achieve guidance for the terminal device 10 do not have a function which acquires information, such as a speed and a moving direction, from the vehicle C10. For this reason, the method of installing the acceleration sensor which measures an acceleration in the terminal device 10, and estimating the position of the terminal device 10 based on the acceleration which the acceleration sensor measured can be considered. For example, a method of performing a stop determination for determining whether the terminal device 10 is moving or stopped based on the acceleration measured by the acceleration sensor can be considered.

  A more specific example will be described. For example, if the terminal device 10 cannot properly receive a signal transmitted from a satellite, the terminal device 10 determines that the vehicle C10 has entered a tunnel or the like, and advances the estimated position as it is at the last specified vehicle speed and traveling direction. And the terminal device 10 determines whether the vehicle C10 has stopped based on the measured acceleration, and when it determines with the vehicle C10 having stopped, it stops the movement of an estimated position.

[1-1. Example of stop judgment technology)
Here, an example of the stop determination technique for determining whether or not the vehicle C10 is stopped will be described. The technique shown here is an example of a technique that is a pre-stage of the present invention, but does not belong to the original conventional technique. In other words, the technique shown here is a technique that the applicant of the present invention has secretly implemented for development, testing, research, and the like, and is not a secret technique such as so-called public, public, or literature.

  For example, as shown in FIG. 1A, the terminal device 10 has each of a short direction of the screen as an x axis, a long direction of the screen as a y axis, and a direction perpendicular to the screen as a z axis. The acceleration in the xyz axis direction is measured. For example, when the terminal device 10 has the screen as the front, the front side is the + z-axis direction, the back side is the -z-axis direction, and when the terminal device 10 is used, the upper side of the screen is the + x-axis direction and the lower side of the screen is − The acceleration in the terminal coordinate system is measured with the x-axis direction, the left side of the screen as the + y-axis direction, and the right side of the screen as the -y-axis direction.

  On the other hand, as shown in FIG. 1B, the moving direction and speed of the vehicle C10 used by the user are such that on the plane perpendicular to the Z axis, the direction in which the vehicle C10 travels is the Z axis. The direction of turning left or right when C10 travels is represented by a vehicle C10 coordinate system in which the Y-axis direction is the Y-axis direction and the vertical direction of the vehicle C10 is the X-axis direction. For example, the moving direction and speed of the vehicle C10 are the + X axis direction for the upward direction of the vehicle C10, the -X axis direction for the downward direction (that is, the ground side), the + Y axis direction for the left turn direction, and the -Y axis for the right turn direction. It is represented by a vehicle coordinate system in which the rear direction of the vehicle C10 is + Z-axis direction and the front direction is -Z-axis direction.

  Further, the terminal device 10 sets the direction of gravity using the acceleration measured in the terminal coordinate system, that is, the −X-axis direction of the vehicle coordinate system, and also uses the acceleration generated when the vehicle C10 moves to use the vehicle C10. Is determined, and a rotation matrix for converting the acceleration measured in the terminal coordinate system into the vehicle coordinate system is obtained based on the estimated reference direction and the estimated traveling direction. Then, the terminal device 10 converts the acceleration in the terminal coordinate system into the acceleration in the vehicle coordinate system using the rotation matrix, and the amplitude, frequency, average value, standard deviation, and maximum value of the converted acceleration in each axial direction. , A total of 18 minimum values are acquired as feature values.

  Further, the terminal device 10 accumulates the feature amount acquired when the speed of the vehicle C10 is equal to or higher than a predetermined threshold value as the feature amount during traveling, and when the speed of the vehicle C10 is equal to or lower than the predetermined threshold value. The acquired feature quantity is accumulated as a stopped feature quantity. Then, the terminal device 10 learns a stop determination model (for example, by SVM (Support Vector Machine) described later) that determines whether or not the vehicle C10 is stopped using the accumulated feature amount, and learns the stop. A determination model is used to determine whether or not the vehicle C10 is stopped when a satellite positioning system such as a tunnel cannot be used.

  However, in such a technique, since the traveling direction of the vehicle C10 is specified using the acceleration generated when the vehicle C10 moves, the vehicle C10 must travel for a certain period in order to obtain the rotation matrix. . As a result, for example, when the vehicle C10 enters the tunnel immediately after starting the guidance, there is a problem that the rotation matrix cannot be obtained and the stop determination cannot be started.

  The user may get off with the terminal device 10 in a service area or the like. As a result, when the attitude of the terminal device 10 changes, the rotation matrix changes, so it is necessary to specify the traveling direction again and obtain the rotation matrix again based on the identified traveling direction and the reference direction. However, even when such processing is executed, the stop determination cannot be started until the traveling direction is specified. In addition, when the road is inclined, an error is likely to occur because a shift occurs between the terminal coordinate system and the vehicle coordinate system.

  A method of calculating an average value of acceleration without converting the acceleration measured in the terminal coordinate system and learning a stop determination model from the feature value of the calculated average value is also conceivable. However, in such a method, since the installation posture of the terminal device 10 is unknown, it is necessary to learn the stop determination model every time guidance is started. Even when such a method is used, if the installation posture of the terminal device 10 changes, it is necessary to learn a stop determination model again, and if the road is inclined, an error may occur. Prone to occur.

  In particular, when the acceleration measured in the terminal coordinate system is coordinate-transformed, the stop determination model can be reused by specifying the rotation matrix corresponding to the traveling direction of the vehicle C10. On the other hand, if the acceleration is not coordinate-transformed, the stop determination model must be learned again when the installation posture of the terminal device 10 changes or whenever guidance is started.

[2. Determination process executed by the terminal device 10 according to the embodiment]
Therefore, the terminal device 10 executes the following determination process. For example, the terminal device 10 detects acceleration in the terminal coordinate system. Further, the terminal device 10 sets a reference direction based on the detected acceleration. And the terminal device 10 acquires the feature-value based on the acceleration of the direction on the basis of the specified reference | standard direction. Then, the terminal device 10 determines the movement state of the terminal device 10 using the acquired feature amount.

  Hereinafter, an example of a functional configuration and an effect of the terminal device 10 that realizes the determination process described above will be described with reference to the drawings.

[2-1. Example of functional configuration)
FIG. 2 is a diagram illustrating an example of a functional configuration of the terminal device according to the embodiment. As illustrated in FIG. 2, the terminal device 10 includes a communication unit 11, a storage unit 12, a plurality of acceleration sensors 13 a to 13 c (hereinafter, may be collectively referred to as “acceleration sensor 13”), and a GPS receiving antenna. 14, an output unit 15, and a control unit 16. The communication unit 11 is realized by, for example, a NIC (Network Interface Card). The communication unit 11 is connected to the network N in a wired or wireless manner. When the communication unit 11 receives the destination from the terminal device 10 and the terminal device 10, the communication unit 11 distributes route information indicating a route to the destination. Send and receive information.

  The storage unit 12 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 12 includes a guidance information database 12a, a learning data database 12b, a test data database 12c, and a model database 12d, which are various data used for executing guidance.

  In the guidance information database 12a, various data used when the terminal device 10 provides guidance is registered. For example, route information to a destination received from a server or the like (not shown) is stored in the guidance information database 12a. The guidance information database 12a stores various images and audio data output in guidance.

  Learning data used for learning the stop determination model is registered in the learning data database 12b. Specifically, feature quantities collected by a predetermined number for each moving speed of the terminal device 10 are registered in the learning data database 12b.

  For example, FIG. 3 is a diagram illustrating an example of information registered in the learning data database according to the embodiment. As illustrated in FIG. 3, information having items such as “state”, “speed zone”, and “data # 1” to “data # 10” is registered in the learning data database 12b. Here, the “state” is information indicating whether the associated learning data is data indicating a running state or data indicating a stopped state. The “speed band” is information indicating the moving speed of the terminal device 10 when the associated feature amount is collected. “Data # 1” to “Data # 10” are feature data.

  For example, in the example illustrated in FIG. 3, when the moving speed of the terminal device 10 is “0 km / h (km / h) to 5 km / h” as learning data indicating “stopped state” in the learning data database 12 b. The collected feature quantity data “feature quantity # 1-1” to “feature quantity # 1-60” are registered. In the learning data database 12b, feature data collected when the moving speed of the terminal device 10 is "5 km / h (km / h) to 15 km / h" as learning data indicating "running state". “Feature amount # 2-1” to “Feature amount # 2-10” are registered.

  In addition, in the learning data database 12b, the moving speed of the terminal device 10 is “15 km / h (km / h) to 25 km / h”, “25 km / h to 35 km / h”, “35 km / h to 45 km”. / H ”,“ 45 km / h to 55 km / h ”,“ 55 km / h or more ”, 10 pieces of feature amount data collected in each speed band are registered.

  That is, in the example shown in FIG. 3, 60 pieces of learning data indicating the running state and 60 pieces of learning data indicating the stopped state are registered in the learning data database 12b. In the example illustrated in FIG. 3, a conceptual value such as “feature amount # 1-1” is described as the data indicating the feature amount, but the embodiment is not limited thereto. As will be apparent from the description to be described later, the learning data database 12b includes, in the learning data database 12b, an average value, standard deviation, maximum value, and minimum value of a set of acceleration magnitude in the reference direction and acceleration magnitude on a plane perpendicular to the reference direction. The value value is registered as data indicating the feature amount.

  In the test data database 12c, feature quantities are registered as test data used for testing a model for determining the movement state of the terminal device 10, that is, a stop determination model. For example, FIG. 4 is a diagram illustrating an example of information registered in the test data database according to the embodiment. As shown in FIG. 4, information having items such as “state”, “test data # 1”, and “test data # 2” is registered in the test data database 12c. Here, the “state” shown in FIG. 4 is information indicating whether the associated test data is data indicating a running state or data indicating a stopped state. “Test data # 1” and “test data # 2” are feature amount data indicating a running state or a stopped state.

  For example, in the example illustrated in FIG. 4, “test data # 1-1” and “test data # 1-2” are registered in the test data database 12c as test data indicating “stopped state”, and “running state”. "Test data # 2-1" and "Test data # 2-2" are registered as test data indicating In addition to the test data shown in FIG. 4, a plurality of test data indicating “stop state” and “running state” (for example, 60 pieces of test data) are registered in the test data database 12c. And In the test data database 12c, it is assumed that more feature quantities than the learning data database 12b are registered as test data.

  The model database 12d holds a plurality of models that determine the movement state of the terminal device 10 when a feature amount based on acceleration is input, and that have different parameters, that is, a plurality of stop determination models. Here, the stop determination model is realized by, for example, a support vector machine (SVM).

  For example, FIG. 5 is a diagram illustrating an example of information registered in the model database according to the embodiment. As shown in FIG. 5, information having items such as “model ID (Identifier)”, “model data”, “correct answer rate (%)”, and “use flag” is registered in the model database 12d. Here, the “model ID” is a model identifier. The “model data” is learned model data.

  The “correct answer rate (%)” is a value indicating the percentage of correct answers when the model is tested using test data. The “use flag” is information indicating a model used by the terminal device 10 as a stop determination model among models registered in the model database 12d. In the example illustrated in FIG. 5, the example in which the model data of the ten models indicated by the model IDs “model # 1” to “model # 10” is stored is described, but the embodiment is limited to this. It is assumed that model data of an arbitrary number of models can be registered.

  For example, in the example shown in FIG. 5, “model data # 1” is registered as model data indicated by “model # 1” in the model database 12d. Is registered that the correct answer rate of the model indicated by "" is "98"%. In the model database 12d, information indicating that the model indicated by “model # 1” among the models indicated by “model # 1” to “model # 10” is used is registered.

  Returning to FIG. 2, the description will be continued. The acceleration sensor 13 measures the magnitude and direction of acceleration related to the terminal device 10 at a predetermined time interval (for example, 5 milliseconds). For example, the acceleration sensor 13a measures the acceleration in the x-axis direction in the terminal coordinate system. The acceleration sensor 13b measures acceleration in the y-axis direction in the terminal coordinate system. The acceleration sensor 13c measures the acceleration in the z-axis direction in the terminal coordinate system. That is, the terminal device 10 can obtain a vector indicating the direction and magnitude of the acceleration with respect to the terminal device 10 by using the acceleration measured by each of the acceleration sensors 13a to 13c as the acceleration in each axial direction of the terminal coordinate system. it can.

  The GPS receiving antenna 14 is an antenna for receiving signals used in a satellite positioning system such as GPS from a satellite. The output unit 15 is a screen for displaying a map and a current location when performing guidance, and a speaker for outputting sound.

  The control unit 16 is stored in a storage device inside the terminal device 10 by, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or the like. The various programs are executed by using a storage area such as a RAM as a work area. In the example illustrated in FIG. 2, the control unit 16 may be described as the guidance execution unit 17, the audio output unit 18, the image output unit 19, and the movement state determination unit 20 (hereinafter collectively referred to as the processing units 17 to 20). .) The movement state determination unit 20 includes a detection unit 21, a setting unit 22, a conversion unit 23, an acquisition unit 24, a determination unit 25, a collection unit 26, a learning unit 27, and an update unit 28.

  The connection relationship between the processing units 17 to 20 included in the control unit 16 is not limited to the connection relationship illustrated in FIG. 2 and may be other connection relationship. Each of the processing units 17 to 20 realizes / executes the function / action (for example, FIG. 1) of the guidance process as described below, but these are functional units arranged for explanation, The actual hardware elements and software modules may be matched. That is, the terminal device 10 may realize and execute the guidance process in arbitrary functional units as long as the following guidance process functions and actions can be realized and executed.

[2-2. Example of effects in guidance processing]
Hereinafter, the contents of the guidance process executed and realized by each of the processing units 17 to 20 will be described using the flowchart shown in FIG. FIG. 6 is a flowchart for explaining an example of a flow of guidance processing executed by the terminal device according to the embodiment.

  First, the guidance execution unit 17 determines whether or not a destination is input from the user (step S101). When the destination is input (step S101: Yes), the guidance execution unit 17 acquires route information from an external server (not shown) (step S102). Here, the guidance execution unit 17 determines whether or not the GPS can be used (step S103).

  For example, the guidance execution unit 17 determines that the GPS cannot be used when the GPS receiving antenna 14 cannot receive a signal from a satellite or when the number of satellites that can receive the signal is smaller than a predetermined threshold. (Step S103: Yes), the current location is acquired from the moving direction and speed of the vehicle C10 estimated by the moving state determination unit 20 (Step S104). For example, the guidance execution unit 17 acquires the current location estimated by the movement state determination unit 20. In addition, the specific content of the process in which the movement state determination part 20 estimates the present location of the vehicle C10 is mentioned later.

  On the other hand, when it is determined that the GPS can be used (step S103: No), the guidance execution unit 17 specifies the current location using the GPS (step S105). Then, the guidance execution unit 17 controls the voice output unit 18 and the image output unit 19 to output guidance using the current location using GPS or the estimated current location (step S106). For example, the voice output unit 18 outputs, from the output unit 15, voice indicating the current location, the direction in which the vehicle C <b> 10 should travel, and the like according to control from the guidance execution unit 17. Further, the image output unit 19 outputs, from the output unit 15, an image in which the current location and the surrounding map are overlaid, an image indicating the direction in which the vehicle C <b> 10 should travel, etc.

  Subsequently, the guidance execution unit 17 determines whether or not the current location is around the destination (step S107). If the guidance execution unit 17 determines that the current location is around the destination (step S107: Yes), the guidance execution unit 17 controls the audio output unit 18 and the image output unit 19 to output end guidance indicating the end of the guidance. (Step S108), and the process ends. On the other hand, when the guidance execution unit 17 determines that the current location is not around the destination (step S107: No), it executes step S103. In addition, when the destination is not input (step S101: No), the guidance execution part 17 waits until it is input.

[2-3. Example of effects in judgment processing]
Next, the contents of the determination process executed and realized by the determination unit 25 will be described with reference to the flowchart shown in FIG. FIG. 7 is a flowchart for explaining an example of a flow of determination processing executed by the terminal device according to the embodiment. For example, if it is determined in step S103 in FIG. 6 that the GPS cannot be used, the determination unit 25 executes the determination process illustrated in FIG. The result of the determination process shown in FIG. 7 is output to the guidance execution unit 17 as the current location estimated by the movement state determination unit 20.

  First, the determination unit 25 is such that the vehicle C10 travels from the position where the GPS can no longer be used at the speed measured when the GPS can no longer be used and in the traveling direction measured when the GPS cannot be used. The current location is estimated (step S201). For example, when the vehicle C10 enters the tunnel, the determination unit 25 estimates the current location on the assumption that the vehicle C10 continues to travel in the speed and direction when the vehicle C10 enters the tunnel. The determination unit 25 may acquire information on a route along which the vehicle C10 is moving from the guidance information database 12a, and may estimate that the vehicle C10 is moving along the acquired route.

  Subsequently, the determination unit 25 reads the model selected from the model database 12d, that is, the model whose use flag is “1” (step S202). Note that the model read by the determination unit 25 is a model that is learned and selected by an acquisition process and a learning process to be described later.

  And the determination part 25 determines whether the vehicle C10 is a stop state using the read model (step S203). Specifically, the determination unit 25 inputs a feature amount acquired by the acquisition unit 24 described later with respect to the read model, and determines whether or not the vehicle C10 is in a stopped state. That is, the determination unit 25 determines the movement state of the terminal device 10 using a model that has learned the features of the feature amount. As described above, the determination unit 25 determines the movement state of the terminal device 10 using the feature amount acquired by the acquisition unit 24.

  And when the determination part 25 determines with the vehicle C10 being in a stop state (step S203: Yes), it stops the movement of the estimated present location (step S204). On the other hand, the determination part 25 performs step S201, when it determines with the vehicle C10 not being in a stop state (step S203: No). For example, if the determination unit 25 determines that the vehicle C10 is in a moving state after determining that the vehicle C10 is in a stopped state, the vehicle C10 is in the speed and direction when the vehicle C10 enters the tunnel. Estimate your current location as you start moving.

[2-4. Example of effects in the acquisition process]
Next, the contents of the acquisition process executed and realized by the detection unit 21, the setting unit 22, the conversion unit 23, and the acquisition unit 24 will be described using the flowchart shown in FIG. FIG. 8 is a flowchart for explaining an example of the flow of acquisition processing executed by the terminal device according to the embodiment. In addition, the detection part 21, the setting part 22, the conversion part 23, and the acquisition part 24 perform the acquisition process shown in FIG. 8 for a predetermined period (for example, every 1 second).

For example, the detection unit 21 acquires acceleration from the acceleration sensor 13 (step S301). Specifically, the acceleration sensor 13 acquires the magnitude of acceleration measured for each (x, y, z) axis direction of the terminal coordinate system at predetermined time intervals. Moreover, the detection part 21 calculates the average value of the magnitude | size of the acceleration which the acceleration sensor 13 measured during the predetermined period for every axial direction of a terminal coordinate system (step S302). For example, the detection unit 21 collects the acceleration in the terminal coordinate system detected by the acceleration sensor 13 every 5 milliseconds for one second. Then, the detection unit 21 calculates an average value x _m of the values of the collected accelerations in the x-axis direction, an average value y _{m of} the values in the y-axis direction, and an average value z _m of the values in the z-axis direction, A vector (x _m , y _m , z _m ) composed of the calculated average values in the respective axis directions is defined as an average vector G.

  Subsequently, the setting unit 22 sets a reference direction based on the acceleration calculated by the detection unit 21 (step S303). More specifically, as illustrated in (C) of FIG. 1, the setting unit 22 sets the direction of the average vector G composed of the average value of accelerations calculated by the detection unit 21 as the reference direction.

  Subsequently, the conversion unit 23 calculates a rotation matrix that matches a predetermined axis direction of the terminal coordinate system with the reference direction set by the setting unit 22 (step S304). And the conversion part 23 converts each component of the acceleration acquired by the detection part 21 by the terminal coordinate system using the calculated rotation matrix (step S305). That is, the conversion unit 23 converts the acceleration acquired by the detection unit 21 into an acceleration in a coordinate system based on the reference direction instead of the vehicle coordinate system.

  For example, as shown in FIG. 1D, the setting unit 22 sets the direction of the average acceleration vector G as a reference. Then, the conversion unit 23 calculates a rotation matrix such that the −x axis direction of the terminal coordinate system and the direction of the average vector G coincide with each other, as shown in FIG. Note that the conversion unit 23 may adopt an arbitrary rotation matrix as long as the −x axis direction of the terminal coordinate system and the direction of the average vector G coincide with each other. That is, the conversion unit 23 may employ a rotation matrix that rotates the y-axis direction or the z-axis direction in an arbitrary direction.

  Then, as shown in FIG. 1 (F), the conversion unit 23 uses the calculated rotation matrix to calculate the acceleration measured in the terminal coordinate system based on the average value of acceleration (hereinafter, estimated). It is described as a coordinate system.) In the following description, the direction of the average vector G in the estimated coordinate system is the −x axis direction. In the following description, the −x axis direction in the estimated coordinate system may be described as the reference direction.

  Subsequently, the acquisition unit 24 calculates the magnitude of the acceleration vector on a plane perpendicular to the direction of the average vector (hereinafter referred to as a horizontal plane), that is, on the yz plane in the estimated coordinate system (step S306). ). Then, the acquisition unit 24 acquires a feature amount from the set of the average vector size and the horizontal acceleration vector size (step S307).

  For example, as illustrated in FIG. 1G, the acquisition unit 24 calculates the acceleration magnitude “a_hor” on the horizontal plane and the average vector G among the axis components of the acceleration coordinate-converted by the conversion unit 23. , That is, a set of acceleration magnitude “a_ver” in the reference direction in the estimated coordinate system. More specifically, when the acceleration component in the estimated coordinate system is described as “a_x, a_y, a_z”, the acquisition unit 24 calculates the square root of the sum of the square of the component “a_y” and the square of the component “a_z”. The value is calculated as “a_hor”.

  Then, as shown in FIG. 1H, the acquisition unit 24 sets the calculated “a_hor” and “a_ver”, that is, the magnitude of the acceleration on the horizontal plane and the magnitude of the acceleration in the reference direction. With respect to “(a_ver, a_hor)” which is a set of and, the average value, the maximum value, the minimum value, and the standard deviation within a predetermined period are acquired as feature amounts. Note that the acquisition unit 24 may use at least one of an average value, a maximum value, a minimum value, and a standard deviation within a predetermined period for “(a_ver, a_hor)” as a feature amount. As described above, the acquisition unit 24 acquires a feature amount based on acceleration in a direction with reference to the reference direction. And the determination part 25 determines the movement state of the terminal device 10 using the feature-value acquired by the acquisition part 24. FIG.

  Thus, the terminal device 10 sets the direction of the average vector as the reference direction for performing stop determination without calculating the direction of gravity or the moving direction of the vehicle C10. Here, the average vector is predicted to indicate the direction of gravitational acceleration when the vehicle C10 is stopped or when the vehicle C10 is moving at a constant speed. For this reason, when the vehicle C10 is stopped or when the vehicle C10 is moving at a constant speed, the reference direction coincides with the direction of gravity acceleration. On the other hand, when the vehicle C10 is accelerating / decelerating, the direction of the average vector G is predicted to deviate from the direction of gravity acceleration.

  However, the terminal device 10 learns a model for performing stop determination using a feature amount based on acceleration, as will be apparent from the description below. More specifically, the terminal device 10 learns a model that performs stop determination based on a feature amount of acceleration when the vehicle C10 is moving and a feature amount of acceleration when the vehicle C10 is stopped. To do. The model that has performed such learning can determine whether or not the vehicle C10 is stopped from the feature amount regardless of the direction of gravity or the moving direction of the vehicle C10. Even when there is an error in the direction of acceleration, the error can be absorbed.

[2-5. Example of effects in learning process]
Next, the contents of the learning process executed and realized by the collection unit 26, the learning unit 27, and the update unit 28 will be described using the flowchart shown in FIG. FIG. 9 is a flowchart illustrating an example of the flow of learning processing executed by the terminal device according to the embodiment. Note that the collection unit 26, the learning unit 27, and the update unit 28 may execute the processing illustrated in FIG. 9 every time a feature amount is acquired by the acquisition unit 24, for example, at predetermined time intervals. May be. Further, the process shown in FIG. 9 is always executed during the execution of the application for realizing the guidance.

  For example, the collection unit 26 collects a predetermined number of feature amounts acquired by the acquisition unit 24 for each moving speed of the terminal device 10 (step S401). More specifically, the collection unit 26 collects a predetermined number of feature amounts in order from the newly acquired feature amount among the feature amounts acquired from b by the acquisition unit 24 for each moving speed of the terminal device 10. To do. Then, the collection unit 26 registers the collected feature amount in the learning data database 12b (Step S402).

  If a specific example is demonstrated, the collection part 26 will specify the moving speed of the terminal device 10 using positioning systems, such as GPS, when the acquisition part 24 acquires the feature-value. Then, the collection unit 26 registers the acquired feature amount in the learning data database 12b as learning data in association with the speed zone including the identified moving speed. Note that the collection unit 26 has an upper limit (for example, “10” in the running state and “60” in the stopped state) associated with the speed band including the specified moving speed. Etc.), the oldest learning data among the associated learning data is updated to the newly acquired feature amount.

  In other words, when a feature quantity whose speed specified using GPS is equal to or less than a predetermined threshold is a feature quantity indicating a stop state and another feature quantity is a feature quantity indicating a running state, the feature quantity is biased. If this exists, the determination accuracy of the model deteriorates. For example, when the model is learned with learning data that occupies most of the feature values collected when the terminal device 10 is moving at a speed of 80 km / h or higher, The determination accuracy is reduced. Therefore, the collection unit 26 collects a predetermined number of feature quantities as learning data in order from the newly acquired feature quantities for each speed zone. As a result, the collection unit 26 can improve the accuracy of the stop determination by the model that has learned using the learning data registered in the learning data database 12 as a result of preventing the speed deviation of the learning data in the running state. .

  In addition, when a single model is simply learned, it cannot be determined whether or not the model is a model that performs an optimal determination. In addition, the parameters of the model that can appropriately perform the stop determination are predicted to change dynamically according to the attitude of the terminal device 10, the specification of the vehicle C10, the movement characteristics of the vehicle C10, and the like.

  Therefore, the learning unit 27 learns a new parameter model using the learning data registered in the learning data database 12b (step S403). Then, the learning unit 27 stores the learned model in the model database 12d. When the vehicle C10 is stopped, there is no vibration from the road surface. Therefore, the value of “a_ver” and the direction of the reference direction substantially coincide with the gravitational acceleration, and the value of “a_hor” becomes almost zero. Conceivable. On the other hand, when the vehicle C10 is traveling at a constant speed, the maximum value, the minimum value, and the standard deviation value of “a_ver” and “a_hor” are stopped by the random vibration from the road surface. It is predicted that it will be different from when you are doing. Further, in the state where the vehicle C10 is accelerating / decelerating, acceleration is added to the acceleration of the vehicle C10 in addition to the acceleration of gravity, so that the magnitude of “a_ver” becomes larger than the acceleration of gravity and “a_hor” Is also estimated to be different from the case where the vehicle C10 is stopped. Therefore, the learning unit 27 can learn a model for determining whether or not the vehicle C10 is stopped based on the average value, the minimum value, the maximum value, and the standard deviation of “(a_ver, a_hor)”. it can.

  Further, the terminal device 10 sets the direction of the average vector as a reference direction for performing stop determination without calculating the direction of gravity or the moving direction of the vehicle C10. As a result, an error occurs between the direction of the average vector G and the direction of gravitational acceleration. As a result, when the vehicle C10 is traveling at a constant speed or when the vehicle C10 is accelerating / decelerating, There is a difference between the feature amount acquired when the vehicle C10 is stopped. For this reason, the terminal device 100 can learn a model that can perform stop determination with higher accuracy.

  Subsequently, the learning unit 27 calculates the correct answer rate of each model using the test data registered in the test data database 12c (step S404), and registers the calculated correct answer rate in the model database 12d. For example, the learning unit 27 inputs all learning data into each model, and when the test data indicating the stop state is input, the stop correct answer rate that is the ratio determined to be the stop state, and the test data indicating the running state And a driving correct answer rate that is a ratio determined to be in the driving state when is input. Then, the learning unit 27 calculates the sum of the value obtained by adding the predetermined coefficient β to the stop correct answer rate and the value obtained by adding the predetermined coefficient α to the running correct answer rate as the correct answer rate. The coefficient α and the coefficient β are coefficients that are set so that the sum thereof is “1”, and are coefficients that are adjusted depending on whether the weight is assigned to the stop determination or the travel determination.

  Subsequently, the learning unit 27 selects the model with the highest correct answer rate from the models registered in the model database 12d (step S405), and sets the use flag associated with the selected model to “1”. To do.

  As a result of such learning processing, the determination unit 25 can determine the movement state of the terminal device 10 by using a predetermined amount of features collected for each movement speed of the terminal device 10. Moreover, the determination part 25 can determine the movement state of the terminal device 10 using the model with the highest correct answer rate among the models held in the model database 12d.

  In addition, the determination unit 25 starts the stop determination using a model whose use flag is “1” among the models registered in the model data 12c at the time of starting an application that realizes the various processes described above. It will be. That is, the terminal device 10 stores information indicating the model that was selected when the application was last activated, and restarts the stop determination using the model indicated by the stored information when the determination is restarted. As a result, the terminal device 10 can eliminate the time period when the stop determination cannot be made.

  Note that the learning unit 27 performs the above-described learning process when the number of learning data indicating the running state and the number of learning data indicating the stopped state become a predetermined number as a result of the collection by the collecting unit 26. It may be executed to generate a model. After that, the learning unit 27 deletes all the data registered in the learning data database 12b, causes the collection unit 26 to collect new learning data, and re-learns the model when the number of learning data is obtained again. May be.

  The collection unit 26 may collect test data. For example, the collection unit 26 registers the characteristics collected when the moving speed of the terminal device 10 is “0” in the test data database 12c as test data indicating a stopped state, and the moving speed of the terminal device 10 is within a predetermined range. The feature amount collected when the vehicle is within (for example, 5 km / h or more) may be registered in the test data database 12c as test data indicating the running state. The collection unit 26 may replace the oldest test data among the test data registered in the test data database 12c with the newly collected feature amount.

  Subsequently, the updating unit 28 updates the model with the lowest correct answer rate among the plurality of models held in the model database 12d to a newly learned model (step S406). For example, the updating unit 28 deletes the model with the lowest correct answer rate from the plurality of models held in the model database 12d, and generates a new model having parameters different from those of the models registered in the model database 12d. Then, the update unit 28 registers a new model in the model database 12.

  Note that the newly registered model is predicted to have a low correct answer rate because the number of times of learning is smaller than that of other models. Therefore, the update unit 28 does not need to update the model when the correct answer rate of the last registered model among the plurality of models stored in the model database 12d is the lowest. By executing such processing, the terminal device 10 can ensure the possibility of improving the accuracy of the stop determination.

  Note that the learning process shown in FIG. 9 may be executed at an arbitrary timing and an arbitrary granularity. For example, the terminal device 10 always executes the processes shown in steps S401 and S402, and when the learning data collection rate is equal to or higher than a predetermined threshold value in each speed band, the process after step S403 may be executed. Good. Further, the terminal device 10 may repeatedly execute the processes after step S403 at a constant cycle.

[3. Example of time until judgment start)
Next, an example of the time until the terminal device 10 starts the stop determination will be described with reference to FIG. FIG. 10 is a diagram illustrating an example of a time until the stop determination is started. For example, as shown in FIG. 10A, in the method using the conventional model, as shown in FIG. 10B, the acceleration measured during the movement of the vehicle C10 for a while from the start. The traveling direction of the vehicle C10 is identified from the above, and the stop determination is started after the traveling direction is identified. For this reason, as shown to (C) in FIG. 10, stop determination cannot be started until the advancing direction of the vehicle C10 is specified.

  On the other hand, as shown in FIG. 10D, in this processing, as shown in FIG. 10E, the −x-axis direction is made to coincide with the direction of the average vector G without specifying the traveling direction. While obtaining the conversion formula periodically, as shown in FIG. 10F, feature quantities (average value of acceleration magnitude, minimum value, maximum value, standard deviation) are calculated with the same period. Then, stop determination is executed using the calculated feature amount. For this reason, the terminal device 10 can start a stop determination rapidly. Further, when the terminal device 10 stores information indicating the model used last time, as shown in FIG. 10 (G), the terminal device 10 uses the previously used model to determine the stop immediately after starting. Can start.

  In this way, the terminal device 10 sets a reference direction based on the average value of acceleration, and acquires a feature amount based on the acceleration in the direction with the reference direction as a reference. And the terminal device 10 performs stop determination using the acquired feature-value. For this reason, the terminal device 10 can immediately execute the stop determination without determining the rotation matrix for specifying the traveling direction and converting the terminal coordinate system to the vehicle coordinate system.

  Further, the terminal device 10 can make a stop determination without obtaining the direction of gravity acceleration and the traveling direction of the vehicle C10. For example, the terminal device 10 converts the terminal coordinate system into an estimated coordinate system based only on the measured acceleration value, that is, a coordinate system that does not depend on a change in the attitude of the terminal device 10. As a result, the terminal device 10 performs coordinate transformation that absorbs the change in posture of the terminal device 10, so that it is possible to realize stop determination that does not depend on the change in posture of the terminal device 10. In particular, since the terminal device 10 periodically obtains the rotation matrix, the terminal device 10 performs stop determination based on a feature amount that depends on the direction of gravity acceleration and the traveling direction of the vehicle C10, that is, a feature amount that requires time for acquisition and identification. Retraining of the model can be eliminated. As a result, the terminal device 10 can determine to stop in a short time after activation or after a change in posture. That is, the terminal device 10 does not have to learn the stop determination model every time the application is used. In addition, the terminal device 10 can restart the stop determination immediately by holding the model used at the time of the previous activation by using the model held at the time of restarting the determination.

[4. Example formula)
Next, an example of processing in which the conversion unit 23 calculates a rotation matrix for converting the terminal coordinate system into the estimated coordinate system will be described using mathematical expressions. In addition, the process which the conversion part 23 performs is not limited to the process which the following numerical formula shows. For example, the conversion unit 23 may perform coordinate conversion from the terminal coordinate system to the estimated coordinate system using a mathematical expression expressing the primary conversion.

For example, each axis of the terminal coordinate system is an xyz axis, and each axis of the estimated coordinate system is an XYZ axis. In such a case, the process of converting the estimated coordinate system into the terminal coordinate system is expressed by the following equation (1). In equation (1), the rotation angle around the x axis is α, the rotation angle around the y axis is β, the rotation angle around the z axis is γ, and the coordinates by rotation around the x axis R _x (α) is a rotation matrix that performs transformation, R _y (β) is a rotation matrix that performs coordinate transformation by rotation around the y axis, and R _z is a rotation matrix that performs coordinate transformation by rotation around the z axis. (Γ).

Further, the rotation matrix R _x (α), the rotation matrix R _y (β), and the rotation matrix R _z (γ) (hereinafter, may be collectively referred to as “each rotation matrix”) are expressed by the following equations. (2) to (4). In the estimated coordinate system, since the −x axis direction only needs to coincide with the direction of the average vector G, an arbitrary value can be set as the value of α.

  Here, since the direction of the average vector G is acceleration in the −X axis direction, it can be expressed by the following expression (5) in the estimated coordinate system.

On the other hand, the average vector G in each axis direction detected in the terminal coordinate system is described as (a _x , a _y , a _z ). In such a case, a _x , a _y , and a _z are values obtained by converting the average vector G shown in Expression (5) using each rotation matrix, and therefore, Expression (6) below is established.

  As a result, Expression (7) is obtained from the value in the z-axis direction in Expression (6).

  Further, considering the size of the average vector G, equation (8) is established, and therefore equation (9) is obtained from the values in the x-axis and y-axis directions in equation (6). As a result, the terminal device 10 can specify the rotation angle β around the y-axis from the equations (7) and (9).

  Here, a positive value is selected as a solution among the values shown in Expression (9). Then, Expression (10) and Expression (11) are obtained from the values in the x-axis and y-axis directions in Expression (6). As a result, the terminal device 10 can specify the rotation angle γ around the z axis from the expressions (10) and (11).

  On the other hand, the process of converting the terminal coordinate system to the estimated coordinate system is the inverse conversion of the coordinate conversion shown in Expression (1), and is expressed by Expression (12) below.

The value of β and γ the formula (7), (9), (10), can be calculated from (11), sample _a x of the acceleration of the terminal coordinate _system, a y, y-axis of _{a z} and z When only the axis is rotated and converted to the estimated coordinate system, equation (13) is obtained. That is, the terminal device 10 converts the terminal coordinate system into the estimated coordinate system using the rotation matrix R _y (β) and the rotation matrix R _z (γ).

[5. (Modification)
The terminal device 10 according to the above-described embodiment may be implemented in various different forms other than the above-described embodiment. Therefore, in the following, another embodiment of the terminal device 10 will be described.

[5-1. (Use general-purpose model)
Here, the terminal apparatus 10 may start an application including a general-purpose model so that the stop determination can be performed from the start time even when the application that realizes the above-described process is started for the first time. When such an application is executed, the terminal device 10 starts the stop determination using a general-purpose model at the time of initial activation, and then uses the model learned for each terminal device 10 to determine the stop. Will be executed.

[5-2. Use of learning model)
The terminal device 10 may determine whether or not the vehicle is stopped based on the magnitude of acceleration on a horizontal plane perpendicular to the reference direction without learning a model for determining stop. For example, the terminal device 10 obtains an average vector G of acceleration when the vehicle is stopped, thereby converting the reference direction into the direction of gravity, and when the magnitude of acceleration on the horizontal plane is smaller than a predetermined threshold value. The vehicle may be determined to be stopped. As described above, the terminal device 10 uses the average vector G of acceleration measured in the terminal coordinate system as a reference direction and determines the traveling state of the vehicle C10 using the feature amount based on the acceleration in the direction based on the reference direction. Even without using a model, the determination may be made in any processing mode.

  Further, the terminal device 10 described above collects learning data for performing model learning. Here, the terminal device 10 may limit the learning data to be collected in order to improve model determination accuracy. For example, the terminal device 10 may discard the feature amount acquired when the speed specified using the GPS is included in the range of 1 km / h to 3 km without using the learning model.

  Moreover, the terminal device 10 may give a weight to a speed band for collecting learning data. For example, the terminal device 10 may increase the number of learning data collected in the “stop state” more than the number of learning data collected in the “running state”. Further, for example, the terminal device 10 increases the number of learning data collected when “5 km / h to 15 km / h” than the number of learning data collected when “55 km / h or more”, The accuracy of separation between the stopped state and the traveling state may be increased.

  Further, the terminal device 10 may set a period for collecting learning data and a period for collecting learning data to be used for each speed band. For example, for the learning data “55 km / h or more”, the terminal device 10 uses the learning data collected up to 5 minutes ago, and for the learning data “25 km / h to 35 km / h”, 12 minutes ago The learning data collected until then may be used. In addition, the terminal device 10 learns a model using learning data before a predetermined time has elapsed since the collection, regardless of the speed band, and learning data after a predetermined time has elapsed since the collection. It may be discarded. The number of data to be collected, the width and number of speed zones, the time zone when the learning data to be collected or used are acquired are not limited by the type of model to be learned, and can be set arbitrarily It is. Moreover, the terminal device 10 may hold a plurality of models learned using learning data collected in different periods.

  In the above description, the terminal device 10 has learned SVM as a model for determining stoppage. However, the embodiment is not limited to this. For example, the terminal device 10 may use learning of a neural network or so-called deep learning. A specific example will be described. The terminal device 10 may learn a neural network such as a DNN (Deep Neural Network) that performs stop determination using the learning data.

  Further, for example, the terminal device 10 may learn a model for estimating the moving speed of the vehicle C10 using the learning data. For example, when the collected learning data is input, the terminal device 10 performs DNN learning so that a node in the output layer corresponding to the speed band from which the learning data is collected reacts. And the terminal device 10 may output the speed zone corresponding to the node of the output layer which reacted when the acquired feature amount was input as the estimation result of the speed zone in which the vehicle C10 moves. Note that the terminal device 10 may perform DNN learning for estimating the moving speed of the vehicle C10 in accordance with the time-series transition of the acquired feature value. In addition, the terminal device 10 learns a plurality of SVMs that determine whether the vehicle C10 is moving in a predetermined speed band using the feature amount, and uses the learned SVMs to determine a speed band in which the vehicle C10 moves. It may be estimated.

  Further, the terminal device 10 may learn a model at an arbitrary timing. For example, the terminal device 10 can hold 180 pieces of learning data, collect 60 pieces of learning data, collect 120 pieces of learning data, and collect 180 pieces of learning data. Re-learn. When the terminal device 10 learns the model using 180 pieces of learning data, the terminal device 10 may discard all the stored learning data and collect new learning data. For safety, the terminal device 10 may determine that the vehicle C10 is always in a stopped state when the model is not learned.

  In addition, when the GPS cannot be used, the terminal device 10 may estimate the moving speed of the vehicle C10 using the last moving speed specified using the GPS and the determination result of the model. For example, when the terminal device 10 enters the tunnel or the like and the GPS becomes unusable, the terminal device 10 calculates the speed specified again using the GPS, that is, the tunnel entry speed that is the speed at the time of entering the tunnel. The current position is advanced using the calculated tunnel approach speed as the previous output speed.

  For example, when the GPS cannot be used, the terminal device 10 selects ten speeds calculated using the GPS so far in the newly calculated order. Then, the terminal device 10 sorts the selected speed values in ascending order, and adopts the eighth speed from the smallest speed as the tunnel entry speed. Here, the terminal device 10 may set the maximum value of the tunnel approach speed.

  Here, when it is determined by the model that the vehicle is stopped, the terminal device 10 determines whether any of the following conditions is satisfied. Specifically, the terminal device 10 determines whether the previous output speed is less than 10 km / h, whether the previous output speed is less than 20 km / h, and whether the previous determination result by the model is in a stopped state. Or whether the previous output speed is less than 30 km / h and whether the previous and previous determination results by the model are in a stopped state. And when either condition is satisfy | filled, the terminal device 10 determines with the vehicle having stopped, and sets last output speed to 0 km / h.

  On the other hand, if none of the conditions is satisfied, the terminal device 10 determines whether or not the determination result of the previous time and the previous time is in a stopped state. If it is determined that the terminal device is in the stopped state, the terminal device 10 Set the value to 1/2. Moreover, the terminal device 10 sets the value of the last output speed to 2/3 when the previous determination result is the stop state. Thereafter, the terminal device 10 determines that the terminal device 10 is moving at the previous output speed.

  That is, when it is determined that the terminal device 10 is in the stopped state, the terminal device 10 is stopped only when the tunnel entry speed is less than 10 km / h, and in other cases, the speed is gradually decreased. On the other hand, when the terminal device 10 is determined to be in the traveling state by the model, the terminal device 10 adds 1/5 of the tunnel approach speed to the previous output speed with the tunnel approach speed as the upper limit.

[5-3. About features)
The terminal device 10 described above uses the average value, the minimum value, the maximum value, and the standard deviation of the magnitude of the acceleration in the direction based on the direction of the average vector G as the feature amount. However, the embodiment is not limited to this. For example, the terminal device 10 may simply use the magnitude of acceleration as the feature amount. For example, the terminal device 10 may use at least one of an average value, a minimum value, a maximum value, and a standard deviation as the feature amount. Alternatively, any combination of these may be used as the feature amount.

  Further, the terminal device 10 described above may change the feature amount used between “a_ver” and “a_hor”. For example, the terminal device 10 may use the minimum value of the “a_ver” value as the feature amount, and may use the standard error of the “a_ver” value as the feature amount depending on the environment. On the other hand, the terminal device 10 may use all of the average value, the minimum value, the maximum value, and the standard deviation of the values of “a_hor” as the feature amount. Further, the terminal device 10 may change the feature amount to be used according to the moving environment of the vehicle C10. That is, the terminal device 10 may appropriately select an effective type of feature amount in order to determine whether or not the vehicle C10 is moving.

[5-4. (About processing interval)
Moreover, the terminal device 10 mentioned above performed the determination process and the learning process at intervals of 1 second, for example. However, the embodiment is not limited to this, and the determination process and the learning process may be executed at an arbitrary timing. For example, the terminal device 10 performs stop determination using GPS, always performs stop determination using a model, and executes the learning process when the correct answer rate becomes lower than a predetermined threshold. Good.

[5-5. Management of posture change)
Here, the terminal device 10 may specify the attitude of the terminal device 10. For example, when the acceleration within a certain period is averaged, the direction of the average vector G coincides with the direction of gravity acceleration. Therefore, the terminal device 10 compares, for example, the average vector direction of all accelerations measured since the application was started and the direction of the average vector of accelerations detected in the last 1 second, When there is an opening of 37 degrees or more in angle (that is, when the cosine value of the angle between each average vector becomes smaller than 0.8), it is determined that the attitude of the terminal device 10 has changed. Also good.

  If it is determined that such a posture change has occurred, the terminal device 10 deletes the learning data registered in the learning data database 12b and each model registered in the model database 12d, and collects new learning data or models. To learn. As a result, the terminal device 10 can reduce the deterioration of the determination accuracy when the posture change occurs.

[5-6. About the stopper function)
Here, when the movement state of the vehicle C10 is determined by the determination process described above, the determination result may deviate from the actual movement state, that is, a guidance runaway state may occur. Therefore, the terminal device 10 may be provided with a stopper function for preventing a runaway state of guidance. For example, the terminal device 10 estimates the current location when a predetermined time (for example, 300 seconds if guidance is provided, 120 seconds if guidance is not provided) has elapsed since the determination of the movement state has started. An elapsed time stopper function for forcibly stopping the movement of the camera may be implemented.

  Further, the terminal device 10 may cancel the function of the elapsed time stopper when entering a specific long tunnel in order to prevent the forced stop by the elapsed time stopper in the tunnel. For example, the terminal device 10 sets a first area at the entrance of a specific tunnel, and releases the elapsed time stopper when the current position of the vehicle C10 overlaps the first area. Moreover, the terminal device 10 may set the second area at the exit of the specific tunnel, and may restart the elapsed time stopper when the current position of the vehicle C10 overlaps the second area.

  Here, when the speed of the vehicle C10 is estimated, when entering the tunnel immediately from the toll gate at the high speed entrance, the tunnel entry speed tends to be set lower than the speed passing through the tunnel. Therefore, the terminal device 10 sets information indicating the speed when passing through the tunnel in the first area set at the position of the tunnel, and the current position of the vehicle C10 overlaps with the first area. The speed indicated by the information set in the first area may be set as the moving speed of the vehicle C10. In addition, the terminal device 10 sets a time until the forced stop by the elapsed time stopper is performed in the first area, and sets the first area when the current position of the vehicle C10 overlaps the first area. The forced stop by the elapsed time stopper may be canceled until the time indicated by the information that has been performed elapses.

  In addition, the terminal device 10 does not enter the tunnel if the GPS does not become unavailable until the predetermined time has elapsed after the current position of the vehicle C10 overlaps the first area. It is not necessary to cancel the elapsed time stopper and to set the moving speed of the vehicle C10. Further, the terminal device 10 sets a speed setting area in the tunnel, and when the current position of the vehicle C10 overlaps with the speed setting area, the terminal device 10 sets the moving speed of the vehicle C10 to the speed indicated by the speed setting area and The time until the forced stop by the time stopper is performed may be reset to a predetermined time. Further, the terminal device 10 may set areas similar to the first area, the second area, the speed setting area, and the like described above as intersections and general tunnels.

[5-7. Other embodiments]
In addition, the said embodiment is only an illustration and this invention includes what is illustrated below and other embodiment other than that. For example, the functional configuration, data structure, processing order and contents shown in the flowchart in the present application are merely examples, and the presence / absence of each element, the order of arrangement and processing execution, and specific contents can be changed as appropriate. . For example, the above-described guidance processing and determination processing can be realized as a device, a method, and a program in a terminal realized by a smartphone application, in addition to the terminal device 10 as exemplified in the above embodiment.

  Moreover, the structure which implement | achieves each process part 17-20 which comprises the terminal device 10 by a further independent apparatus is also common. Moreover, the structure which implement | achieves each part 21-25 which comprises the movement state determination part 20 with an respectively independent apparatus may be sufficient. Similarly, the configuration of the present invention is flexible, such as by realizing each means shown in the above embodiment by calling an external platform or the like with an API (application program interface) or network computing (so-called cloud). Can be changed. Furthermore, each element such as means relating to the present invention may be realized by other information processing mechanisms such as a physical electronic circuit as well as a computer control unit.

  For example, the terminal device 10 may execute the guidance process described above in cooperation with a distribution server that can communicate with the terminal device 10. For example, the distribution server includes a collection unit 26, a learning unit 27, and an update unit 28, collects feature amounts from acceleration detected by the terminal device 10, learns a model using the collected feature amounts, and learns The model may be distributed to the terminal device 10. In addition, the distribution server includes a detection unit 21, a setting unit 22, a conversion unit 23, an acquisition unit 24, and a determination unit 25. Based on the acceleration value detected by the terminal device 10, the distribution server calculates the estimated movement direction and movement speed. You may deliver to the terminal device 10 and may perform a user's guidance. Further, the distribution server may cause the terminal device 10 to execute the guidance process by executing the determination process described above instead of the terminal device 10 and transmitting the execution result to the terminal device 10.

  When there are a plurality of terminal devices that perform guidance processing and determination processing in cooperation with the distribution server, the distribution server uses different SVMs for each terminal device to determine whether each terminal device is moving. You may judge. In addition, the distribution server collects position information acquired by each terminal device using GPS, determines whether each terminal device is moving from the collected position information, and collects the determination result and each terminal device. SVM learning may be realized using the acceleration value.

[6. effect〕
As described above, the terminal device 10 detects acceleration, sets a reference direction based on the detected acceleration, acquires a feature amount based on acceleration in a direction based on the set reference direction, and is acquired. The moving state of the terminal device 10 is determined using the feature amount. For this reason, the terminal device 10 can easily determine the traveling state regardless of the installation posture of the terminal device 10. That is, since the terminal device 10 performs coordinate conversion that can absorb the change in posture without specifying the traveling direction or the gravity direction of the vehicle C10, the traveling state can be easily changed regardless of the installation posture of the terminal device 10. Can be judged.

  Moreover, the terminal device 10 sets a reference direction using the average value of the detected acceleration. Therefore, the terminal device 10 can easily determine the traveling state by setting the reference direction regardless of the installation posture of the terminal device 10.

  Further, the terminal device 10 converts the detected acceleration into an acceleration in a direction based on the reference direction using a rotation matrix that matches a predetermined axial direction for detecting the acceleration with the set reference direction, and after the conversion The feature amount based on the acceleration of the is acquired. For this reason, the terminal device 10 can determine the moving state of the vehicle C10 based on versatile feature amounts that do not depend on the installation posture of the terminal device 10 or the moving direction of the vehicle C10.

  In addition, the terminal device 10 acquires a feature amount based on the magnitude of acceleration in a direction based on the reference direction. In addition, the terminal device 10 acquires a feature amount based on acceleration on an estimated plane perpendicular to the reference direction. More specifically, the terminal device 10 acquires a feature amount based on a set of a magnitude of acceleration in the reference direction and a magnitude of acceleration in a direction (for example, on a horizontal plane) with the reference direction as a reference. For this reason, the terminal device 100 can determine the moving state of the vehicle C10 based on the feature amount that does not depend on the installation posture of the terminal device 10 or the moving direction of the vehicle C10.

  Further, the terminal device 10 determines the movement state of the terminal device 10 based on at least one of the average value, maximum value, minimum value, and standard deviation of the magnitude of acceleration. Therefore, the terminal device 10 can easily determine the moving state of the vehicle C10 regardless of whether the vehicle is stopped, moving, or accelerating / decelerating.

  In addition, the terminal device 10 determines the movement state of the terminal device 10 using a model that has learned the features of the feature amount. That is, the terminal device 10 determines the moving state of the vehicle C10 using a model that does not depend on the attitude change of the terminal device 10. For this reason, the terminal device 10 can reduce the time taken for starting the movement determination, for example.

  Further, the terminal device 10 collects a predetermined number of acquired feature values for each moving speed of the terminal device 10, and uses the feature values collected by a predetermined number for each moving speed of the terminal device 10, The movement state of the terminal device 10 is determined. For this reason, the terminal device 10 can prevent the deterioration of the determination accuracy caused by the deviation of the moving speed.

  In addition, the terminal device 10 collects a predetermined number of feature amounts in order from the newly acquired feature amount among the acquired feature amounts for each moving speed of the terminal device 10. For this reason, the terminal device 10 can gradually improve the determination accuracy in each speed band.

  In addition, the terminal device 10 holds a plurality of models for determining the movement state learned based on the feature amount. And the terminal device 10 determines a movement state using the model with the highest correct answer rate among the held models. Moreover, the terminal device 10 updates the model with the lowest correct answer rate among the plurality of held models to a newly learned model. For this reason, the terminal device 10 can determine the moving state of the vehicle C10 with high accuracy.

  Further, when the determination of the movement state of the terminal device 10 is stopped, the terminal device 10 stores a flag indicating the model used. And when resuming the determination of the movement state of the terminal device 10, the terminal device 10 determines the movement state of the terminal device 10 using the model which the stored flag shows. For this reason, the terminal device 10 can start the determination of the moving state of the vehicle C10 after starting the application for determining the moving state without learning the model or specifying the moving direction of the vehicle C10. .

  As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to implement the present invention in other forms with improvements.

  Moreover, the above-mentioned “section (module, unit)” can be read as “means”, “circuit”, and the like. For example, the movement state determination unit can be read as movement state determination means or a movement state determination circuit.

DESCRIPTION OF SYMBOLS 11 Communication part 12 Memory | storage part 13 Acceleration sensor 14 GPS receiving antenna 15 Output part 16 Control part 17 Guidance execution part 18 Audio | voice output part 19 Image output part 20 Moving state determination part 21 Detection part 22 Setting part 23 Conversion part 24 Acquisition part 25 Determination Part 26 collecting part 27 learning part 28 updating part

Claims (21)

A detection unit for detecting acceleration;
A setting unit for setting a reference direction based on the acceleration detected by the detection unit at a preset time interval;
An acquisition unit that acquires a feature amount based on acceleration in a direction based on the reference direction set by the setting unit;
A determination unit that determines a moving state of the terminal device using the feature amount acquired by the acquisition unit ;
The determination unit, wherein the acquisition unit acquires the feature amount based on a set of a magnitude of acceleration in the reference direction and a magnitude of acceleration in a direction based on the reference direction . A detection unit for detecting acceleration;
A setting unit for setting a reference direction based on the acceleration detected by the detection unit at a preset time interval;
An acquisition unit that acquires a feature amount based on acceleration in a direction based on the reference direction set by the setting unit;
A determination unit that determines a movement state of the terminal device using the feature amount acquired by the acquisition unit;
A collection unit that collects a predetermined number of feature quantities acquired by the acquisition unit for each moving speed of the terminal device;
The determination unit determines a moving state of the terminal device using a feature amount collected by a predetermined number for each moving speed of the terminal device.
A determination apparatus characterized by that. A detection unit for detecting acceleration;
A setting unit for setting a reference direction based on the acceleration detected by the detection unit at a preset time interval;
An acquisition unit that acquires a feature amount based on acceleration in a direction based on the reference direction set by the setting unit;
A determination unit that determines a movement state of the terminal device using the feature amount acquired by the acquisition unit;
Holding a plurality of models for determining the movement state learned based on the feature amount,
The determination unit determines the movement state using a model having the highest correct answer rate among the models held by the holding unit.
A determination apparatus characterized by that. The said acquisition part acquires the said feature-value based on the group of the magnitude | size of the acceleration of the said reference direction, and the magnitude of the acceleration of the direction on the basis of the said reference direction. The Claim 2 or 3 characterized by the above-mentioned. Judgment device. A collection unit that collects a predetermined number of feature amounts acquired by the acquisition unit for each moving speed of the terminal device;
The determination unit uses the characteristic amount that is collected by a predetermined number for each moving speed of the terminal device, determining device according to claim 1 or 3, wherein the determining the moving state of the terminal device . A holding unit that holds a plurality of models for determining the movement state learned based on the feature amount;
The determination unit, of the model held by the holding unit, using the highest model correct rate, determination apparatus according to claim 1 or 2, characterized in that determining said moving state. The setting unit is configured by using the average value of the detected acceleration by the detection unit, the determination device according to any one of claims 1 to 6, characterized in that to set the reference direction. The acquisition unit uses the rotation matrix that matches a predetermined axial direction in which the detection unit detects acceleration as a reference direction set by the setting unit, and uses the reference direction for the acceleration detected by the detection unit. and then was converted to the direction of the acceleration, the determination device according to any one of claims 1-7, characterized in that to obtain the characteristic amount based on the acceleration after conversion. The acquisition unit, the determination device according to any one of claims 1-8, characterized in that to obtain the characteristic amount based on the magnitude of the acceleration in the directions with reference to the reference direction. The acquisition unit, according to any one of claims 1-9, characterized in that to obtain the characteristic amount based on the acceleration of a plane perpendicular to the reference direction set by the setting unit Judgment device. The determination unit determines a moving state of the terminal device based on at least one of an average value, a maximum value, a minimum value, and a standard deviation of the magnitude of the acceleration as the feature amount. The determination device according to any one of claims 1 to 10 . The determination device according to any one of claims 1 to 11 , wherein the determination unit determines a movement state of the terminal device using a model in which a feature of the feature amount is learned. The said collection part collects a predetermined number of feature-values for every moving speed of the said terminal device in order from the newly acquired feature-value among the feature-values acquired from the said acquisition part. 6. The determination device according to 2 or 5 . The determination apparatus according to claim 3, further comprising: an update unit that updates a model having the lowest correct answer rate to a newly learned model among the plurality of models held in the holding unit. When the determination of the movement state of the terminal device by the determination unit is stopped, the storage unit stores information indicating a model used by the determination unit;
The determination unit, when restarting judgment of the moving state of the terminal apparatus, using a model indicated by the information the storage unit stores, claim 3, wherein the determining the moving state of the terminal device , 6 or 14 . A determination method executed by a determination device,
A detection process for detecting acceleration;
A setting step for setting a reference direction based on the acceleration detected by the detection step at a preset time interval;
An acquisition step of acquiring a feature amount based on an acceleration in a direction based on the reference direction set in the setting step;
A determination step of determining a movement state of the terminal device using the feature amount acquired by the acquisition step ,
In the acquisition step, the feature amount is acquired based on a set of a magnitude of acceleration in the reference direction and a magnitude of acceleration in a direction with reference to the reference direction . A determination method executed by a determination device,
A detection process for detecting acceleration;
A setting step for setting a reference direction based on the acceleration detected by the detection step at a preset time interval;
An acquisition step of acquiring a feature amount based on an acceleration in a direction based on the reference direction set in the setting step;
A determination step of determining a moving state of the terminal device using the feature amount acquired by the acquisition step;
Collecting the feature amount acquired by the acquisition step by a predetermined number for each moving speed of the terminal device, and
In the determination step, the moving state of the terminal device is determined using a feature quantity collected by a predetermined number for each moving speed of the terminal device.
The determination method characterized by this. A determination method executed by a determination device,
A detection process for detecting acceleration;
A setting step for setting a reference direction based on the acceleration detected by the detection step at a preset time interval;
An acquisition step of acquiring a feature amount based on an acceleration in a direction based on the reference direction set in the setting step;
A determination step of determining a moving state of the terminal device using the feature amount acquired by the acquisition step;
Holding a plurality of models for determining the movement state learned based on the feature amount, and
The determination step determines the movement state using a model having the highest correct answer rate among the models held in the holding step.
The determination method characterized by this. Computer
Detection means for detecting acceleration;
Setting means for setting a reference direction based on the acceleration detected by the detection means at a preset time interval;
Obtaining means for obtaining a feature quantity based on acceleration in a direction based on the reference direction set by the setting means;
Using the feature amount acquired by the acquisition unit, function as a determination unit that determines the movement state of the terminal device ,
The determination program characterized in that the acquisition means acquires the feature quantity based on a set of a magnitude of acceleration in the reference direction and a magnitude of acceleration in a direction based on the reference direction . Computer
Detection means for detecting acceleration;
Setting means for setting a reference direction based on the acceleration detected by the detection means at a preset time interval;
Obtaining means for obtaining a feature quantity based on acceleration in a direction based on the reference direction set by the setting means;
Determination means for determining a movement state of the terminal device using the feature amount acquired by the acquisition means;
Collecting the feature amount acquired by the acquisition unit as a collection unit that collects a predetermined number for each moving speed of the terminal device;
The determination unit determines a moving state of the terminal device using a feature amount collected by a predetermined number for each moving speed of the terminal device.
Judgment program characterized by that. Computer
Detection means for detecting acceleration;
Setting means for setting a reference direction based on the acceleration detected by the detection means at a preset time interval;
Obtaining means for obtaining a feature quantity based on acceleration in a direction based on the reference direction set by the setting means;
Determination means for determining a movement state of the terminal device using the feature amount acquired by the acquisition means;
A holding unit that holds a plurality of models for determining the movement state learned based on the feature amount;
The determination means determines the movement state using a model having the highest correct answer rate among the models held by the holding means.
Judgment program characterized by that.

Images