JP5826049B2 - Moving vehicle estimation method, mobile terminal, and program for estimating moving vehicle on which user is on board - Google Patents

Description

  The present invention relates to a technique for estimating a traveling route based on a current position acquired by a positioning unit such as a GPS (Global Positioning System) mounted on a mobile terminal.

  Various services based on the current location are provided for mobile terminals such as mobile phones and smartphones. The traveling route can be estimated by the mobile terminal always positioning the current position. A user who operates the mobile terminal can receive provision of a service in real time according to the travel route. Here, GPS is generally used as a positioning function for the current position.

  Conventionally, there is a technique for estimating a three-dimensional movement path using sensor (acceleration sensor, angular velocity sensor, atmospheric sensor, etc.) information in addition to a positioning unit that measures the current position (see, for example, Patent Document 1). According to this technique, a GPS positioning unit is used for estimating a two-dimensional movement route.

  In addition, there is a technique for extracting a walking section and a non-walking section based on a walking state, and estimating a moving means (train, bus, car) from a moving speed based on position measurement and time measurement for the non-walking section (for example, Patent Document 2). According to this technique, even if there is an error in the moving speed and the moving speed is not constant, the determination accuracy of the moving means can be improved.

  Furthermore, there is a map matching technique that corrects the current position measured by collating the current position measured by the moving vehicle with road data (see, for example, Patent Document 3). According to this technique, even if the straight running is continued for a long time, the accumulated error does not increase, and the accuracy of map matching can be prevented from being lowered. Similarly, as a technique for map matching, there is a technique for calculating a candidate for a traveling route on which a vehicle is traveling from a road map according to a travel distance in units of periods (see, for example, Patent Document 4).

JP 2002-48589 A JP 2010-1112750 A JP-A-8-334344 JP-A-8-334371 JP-A-8-334369 JP 2003-344068 A JP 2010-286344 A

  According to the technique described in Patent Document 1, pattern matching in the power spectrum of the acceleration sensor is used, but sudden noise and vibration pattern fluctuations are not considered in the estimation of the movement path. In particular, when the user having the mobile terminal is on an electric vehicle, there is a possibility that the estimation of the movement route is erroneous. In addition, since the map information is not taken into consideration, there is room for improvement in the position estimation accuracy.

  Moreover, according to the technique described in Patent Document 2, in order to estimate a moving vehicle based on the maximum speed and maximum acceleration, the estimation may be erroneous depending on factors such as various road congestion conditions and railway operation conditions. There is also sex. In addition, since the map information is not taken into consideration, there is room for improvement in the position estimation accuracy.

  Furthermore, according to the technique described in Patent Document 3, it is a map matching technique specialized for automobiles, and cannot be applied to other moving vehicles such as railways. Further, according to the technique described in Patent Document 4, the optimum map matching is not applied to each moving vehicle (for example, an automobile, a railway train, a bus, etc.).

  Therefore, the present invention provides a moving vehicle estimation method, a mobile terminal, and a program capable of estimating what type of moving vehicle is boarded among the electric vehicles on which the user carrying the mobile terminal is boarding. The purpose is to provide.

According to the present invention, a mobile vehicle estimation method for a mobile terminal having a positioning unit for positioning by receiving positioning radio waves,
The mobile terminal
A first map accumulating unit that accumulates a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating unit that accumulates a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. An accumulation unit;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. A storage unit,
The mobile terminal
A first step of storing the current position measured by the positioning unit as time elapses;
A second step of calculating a first error between the current position and a node of the first map, and calculating a second error between the current position and a node of the second map;
A third step of calculating at least a one-dimensional first current vector having the first error accumulation value as an element and calculating at least a one-dimensional second current vector having the second error accumulation value as an element; When,
The first a posteriori probability of boarding the first moving vehicle is calculated on the condition of the first current vector based on the first learning vector, and the second current vector based on the second learning vector is used as a condition. A fourth step of calculating a second posterior probability of boarding the second moving vehicle;
Fifth step of comparing the first posterior probability with the second posterior probability and estimating that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by having.

According to the onset bright, a mobile vehicle estimating method of a mobile terminal having a positioning portion for positioning the position by receiving the positioning radio wave,
The mobile terminal
A first map accumulating unit that accumulates a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating unit that accumulates a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulating unit that accumulates a first learning vector including a degree of coincidence between a past travel position accumulated for learning and a first map travel route for the first moving vehicle;
A second learning vector accumulating unit that accumulates a second learning vector including a degree of coincidence between a past travel position accumulated for learning and a second map travel route for the second moving vehicle;
The mobile terminal
A first step of storing the current position measured by the positioning unit as time elapses;
A second step of calculating a first degree of coincidence between the current position and the route of the first map, and calculating a second degree of coincidence between the current position and the route of the second map;
A third step of calculating a first current vector including a first degree of match and calculating a second current vector including a second degree of match;
The first a posteriori probability of boarding the first moving vehicle is calculated on the condition of the first current vector based on the first learning vector, and the second current vector based on the second learning vector is used as a condition. A fourth step of calculating a second posterior probability of boarding the second moving vehicle;
Fifth step of comparing the first posterior probability with the second posterior probability and estimating that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by having.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
The mobile terminal further includes an acceleration sensor,
It is also preferable to further include a step of estimating a moving state as to whether or not the vehicle is “onboard an electric vehicle” based on acceleration data output from the acceleration sensor as a first stage of the first step.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
The estimation of the movement state is a case where it is estimated that the first movement state-> second movement state-> first movement state as time elapses, and the second movement state is within a predetermined time. In some cases, it is also preferable to estimate these movement states as one first movement.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
The first moving vehicle is a road vehicle;
The first map is a road route, with the corners that cross the intersection as nodes,
The second mobile vehicle is a rail vehicle,
It is also preferable for the second map to have a railway route and a station as a node.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
For three or more moving vehicles, each moving vehicle has a map storage unit and a learning vector storage unit,
For the second to fourth steps , calculate for each moving vehicle ,
Regarding the fifth step, it is also preferable to compare the posterior probabilities for each moving vehicle and estimate that the user holding the mobile terminal is on the moving vehicle corresponding to the current vector having the highest posterior probability.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
Mobile vehicles also preferably include railroad trains, cars and buses.

According to another embodiment of the mobile vehicle estimation method of the mobile terminal of the present invention,
The current vector and the learning vector are other elements:
Estimated speed = path length / time required,
Average of the electric field strength between the base station,
Dispersion of electric field strength between base stations,
Average stop interval in movement and / or
It is also preferable to further include dispersion of stop intervals in the movement.

According to the present invention, there is a mobile terminal that has a positioning unit that measures a position by receiving positioning radio waves, and estimates a moving vehicle on which a user carrying the mobile terminal is boarded,
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. Storage means;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. Storage means;
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first error between the current position and a node of the first map;
A second map matching means for calculating a second error between the current position and a node of the second map;
First current vector calculation means for calculating at least a one-dimensional first current vector having the first accumulated error value as an element ;
Second current vector calculation means for calculating at least a one-dimensional second current vector having the second error accumulated value as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
The moving vehicle estimation means for comparing the first posterior probability with the second posterior probability and estimating that the user having the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by having.

According to the present invention, there is a mobile terminal that has a positioning unit that measures a position by receiving positioning radio waves, and estimates a moving vehicle on which a user carrying the mobile terminal is boarded,
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulation in which at least a one-dimensional first learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the first map as an element is accumulated for the first moving vehicle. Means,
Second learning vector accumulation in which at least a one-dimensional second learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the second map as an element is accumulated for the second moving vehicle. Means,
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first matching degree between the current position and the route of the first map;
A second map matching means for calculating a second matching degree between the current position and the route of the second map;
First current vector calculation means for calculating a first current vector of at least one dimension having a first degree of match as an element ;
Second current vector calculating means for calculating at least one-dimensional second current vector having the second degree of match as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
The moving vehicle estimation means for comparing the first posterior probability with the second posterior probability and estimating that the user having the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by having.

According to the present invention, a computer mounted on a terminal having a positioning unit that measures a position by receiving positioning radio waves and an acceleration sensor is used to estimate a moving vehicle on which a user having the mobile terminal is boarded. A program that functions to
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. Storage means;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. Storage means;
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first error between the current position and a node of the first map;
A second map matching means for calculating a second error between the current position and a node of the second map;
First current vector calculation means for calculating at least a one-dimensional first current vector having the first accumulated error value as an element ;
Second current vector calculation means for calculating at least a one-dimensional second current vector having the second error accumulated value as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition Second posterior probability calculating means for calculating a second posterior probability of boarding the second moving vehicle;
The moving vehicle estimation means for comparing the first posterior probability with the second posterior probability and estimating that the user having the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by functioning a computer.

According to the present invention, a computer mounted on a terminal having a positioning unit that measures a position by receiving positioning radio waves and an acceleration sensor is used to estimate a moving vehicle on which a user having the mobile terminal is boarded. A program that functions to
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulation in which at least a one-dimensional first learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the first map as an element is accumulated for the first moving vehicle. Means,
Second learning vector accumulation in which at least a one-dimensional second learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the second map as an element is accumulated for the second moving vehicle. Means,
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first matching degree between the current position and the route of the first map;
A second map matching means for calculating a second matching degree between the current position and the route of the second map;
First current vector calculation means for calculating a first current vector of at least one dimension having a first degree of match as an element ;
Second current vector calculating means for calculating at least one-dimensional second current vector having the second degree of match as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
The moving vehicle estimation means for comparing the first posterior probability with the second posterior probability and estimating that the user having the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. It is characterized by functioning a computer.

  According to the mobile vehicle estimation method, mobile terminal, and program of the present invention, it is possible to estimate which mobile vehicle is on board among the electric vehicles on which the user carrying the mobile terminal is on board. . Thereby, the estimation accuracy of the waypoint of the traveling route can be improved.

It is a function block diagram of the mobile terminal in this invention. It is explanatory drawing showing the difference | error of the present position and a traffic node. It is explanatory drawing showing the coincidence degree of the present position and a traffic route. It is a function block diagram of the database which calculates a learning vector. It is explanatory drawing showing calculation of the posterior probability for every moving vehicle. It is explanatory drawing which derives | leads-out a power spectrum from acceleration data. It is explanatory drawing which derives | leads-out the 1st probability model of a power spectrum. It is explanatory drawing which derives | leads-out the 2nd probability model in consideration of the confidence interval. It is explanatory drawing which derives | leads-out a confidence interval.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a functional configuration diagram of a mobile terminal according to the present invention.

  A mobile terminal 1 such as a mobile phone or a smartphone is carried by a user and moves with the user. According to FIG. 1, the mobile terminal 1 includes a positioning unit 101 that measures the current position, a wide area communication interface unit 102, and an acceleration sensor 103.

  The positioning unit 101 receives the positioning radio wave from the GPS satellite 2 to calculate and output the latitude / longitude information of the current position.

  The wide area communication interface unit 102 has, for example, a mobile phone communication function and can be connected to the Internet via a base station. Thereby, the mobile terminal 1 can communicate with the database via the Internet. The wide area communication interface unit 102 can also measure the electric field strength of radio waves from the base station.

  The acceleration sensor 103 is an inertial sensor for measuring acceleration, and can detect a change in velocity per unit time (acceleration of a DC component). The triaxial acceleration sensor is mounted on a general mobile phone. A user who has a mobile phone vibrates his / her body while walking to detect a change in the speed of vertical / horizontal motion.

  In addition, according to FIG. 1, the mobile terminal 1 includes a railway line accumulation unit 110, a road map accumulation part 120, a bus route accumulation part 130, a railway line learning vector accumulation part 111, and a road map learning vector accumulation part 121. And a bus route learning vector storage unit 131.

The railway line accumulation unit 110 accumulates an railway line (first map) including a passage route (track) and a passage node (for example, a station) in the railway train (first moving vehicle).
The road map accumulation unit 120 accumulates a road map (second map) including a traffic route (road) and a traffic node (for example, a corner or an intersection) in an automobile (second moving vehicle).
The bus route accumulation unit 130 accumulates a bus route (third map) including a traffic route (road) and a traffic node (stop) in the bus (third moving vehicle).
In addition, a traffic route consists of a continuation of several latitude longitude information, and a traffic node consists of one latitude longitude information.

The railway line learning vector accumulating unit 111 accumulates errors between past passage positions accumulated for learning and passage nodes (for example, stations) of the railway line (first map) for the railway train (first moving vehicle). The train line learning vector including the value is accumulated.
The road map learning vector accumulating unit 121 has an error between a past traffic position accumulated for learning and a traffic node (for example, a corner or an intersection) of the road map (second map) for the automobile (second moving vehicle). Accumulate road map learning vectors including cumulative values.
The bus route learning vector accumulation unit 131 accumulates an error between a past passage position accumulated for learning and a passage node (for example, a stop) of the bus route (third map) for the bus (third moving vehicle). The bus route learning vector including is stored.
The learning vector will be described later with reference to FIG.

  Further, according to FIG. 1, the mobile terminal 1 includes a position accumulation storage unit 141, a railway line map matching unit 112, a road map map matching unit 122, a bus route map matching unit 132, and a railway current vector calculation unit 113. A vehicle current vector calculation unit 123, a bus current vector calculation unit 133, a moving state estimation unit 142, a posterior probability calculation unit 143, a moving vehicle estimation unit 144, and an application unit 145. These functional components are realized by executing a program that causes a computer mounted on the mobile terminal to function.

  The position accumulation storage unit 141 cumulatively stores the positions measured by the positioning unit 101 as time elapses. Accumulate latitude and longitude according to time. The accumulated cumulative position information is output to the map matching units 112, 122, and 132.

Here, the following two values can be applied as the map matching.
・ Error between current position and traffic node ・ Matching degree between current position and route

[Error between current position and traffic node]
The railway line map matching unit 112 calculates a first error between the current position and the travel node (for example, a station) of the railway line (first map).
The road map map matching unit 122 calculates a second error between the current position and a road node (for example, a corner or an intersection) of the road map (second map).
The bus route map matching unit 132 calculates a third error between the current position and a traffic node (for example, a stop) of the road map (third map).

  FIG. 2 is an explanatory diagram showing an error between the current position and the travel node.

  According to FIG. 2, railway lines, roads, and bus routes are represented. Each route represents a traffic node. “Traffic node” is, for example, a station in the case of a railway, a corner or an intersection in the case of a road, and a stop in the case of a bus. Then, the “current position” is associated with the nearest “passage node” in each route. For example, only the current position when the distance to the nearest traffic node is closest may be extracted. Also, instead of a traffic node, even if a perpendicular line is dropped to a line segment connected by a traffic node on a railroad, road, or bus route, a perpendicular foot (intersection of the vertical line and the line segment) is used. Good.

The railway current vector calculation unit 113 calculates a railway (first) current vector including the first accumulated error value.
The vehicle current vector calculation unit 123 calculates a vehicle (second) current vector including the second error accumulation value.
The bus current vector calculation unit 133 calculates a bus (third) current vector including the third accumulated error value.
The current vector has at least a one-dimensional error accumulation value. Here, it is also preferable to further include other feature elements (dimensions) in the current vector.
Current vector (F1, F2, F3, F4, F5, F6, F7)
F1: Cumulative error E between the current position and the traffic node
F2: estimated speed = path length / required time F3: electric field strength average F4: electric field strength variance F5: stop interval average F6: stop interval variance F7: moving vehicle probability value

(F1) As for the error accumulation value, the value in the correct passage route becomes smaller. In particular, the difference between the current position and the correct passing node is considered to be small.

Note that F2 to F7 are optionally included in the current vector.
(F2) The estimated speed is generally considered that the railway train is faster than the automobile and the automobile is faster than the bus.
(F3, F4) The electric field strength is the strength of the radio wave transmitted from the base station. For example, when the railway is a subway, the electric field strength is considered to be extremely small.
(F5, F6) The stop interval is a time interval of temporary stop. In general, a railway train has a longer stop interval than an automobile, and an automobile has a longer stop interval than a bus.
(F7) The values in the probability table estimated by the detailed classification (electric train, automobile, bus) of the electric vehicle for the movement state estimation unit 142. It is optional and takes into account the results in the movement state estimation process.
Note that the feature elements F3 to F6 have only to be calculated once because they are common feature amounts for each moving vehicle.

[Matching degree between current position and route]
The railway line map matching unit 112 calculates a first degree of coincidence between the current position and the route (for example, a track) of the railway line (first map).
The road map map matching unit 122 calculates a second degree of coincidence between the current position and a road route (for example, a road) of the road map (second map).
The bus route map matching unit 132 calculates a third degree of coincidence between the current position and the road route (for example, road) of the road map (third map).

  FIG. 3 is an explanatory diagram showing the degree of coincidence between the current position and the travel route. It should be noted that the calculation method is the same for all routes of railway lines, roads, and bus routes.

According to FIG. 3, the first traffic route and the second traffic route are represented. In order to associate the current position with the traffic route, a route search is performed using the current location by the following two methods to derive a first traffic route and a second traffic route.
(First traffic route) Traffic that connects the search results of these routes with the current position (in the case of FIG. 3, P1 and P2, P2 and P3, P3 and P4 in FIG. 3) as the starting point and destination, respectively. A route is derived.
(Second travel route) The start route of the current position (P1 in the case of FIG. 3) and the end point of the current position (P4 in the case of FIG. 3) are set as the starting point and the destination, respectively, and the route of the search result of the route Is derived. For route search when deriving the first and second traffic routes, a route search that matches as much as possible to the time from the current position as the starting point to the current position as the destination from among a plurality of search results. It is desirable to choose.

Here, “the degree of match between the current position and the travel route” is the path length or time length for which the first route search result and the second route search result match each other for the first route. Divided by the path length or time length of the search result. In the case of FIG. 3, the matching degree is as follows.
Matching degree when using path length = (a + e) / (a + b + c + d + e)
Consistency with time length = (s + x) / (s + t + u + w + x)

The railway current vector calculation unit 113 calculates a railway (first) current vector including the first matching degree.
The vehicle current vector calculation unit 123 calculates a vehicle (second) current vector including the second degree of match.
The bus current vector calculation unit 133 calculates a bus (third) current vector including the third matching degree.
The current vector has at least a one-dimensional match value cumulative value. Here, it is also preferable to further include other feature elements (dimensions) in the current vector.
Current vector (F1, F2, F3, F4, F5, F6, F7)
F1: Match C between the current position and the route
F2: estimated speed = path length / required time F3: electric field strength average F4: electric field strength variance F5: stop interval average F6: stop interval variance F7: moving vehicle probability value

(F1) With respect to the degree of match between the current position and the route, the value on the correct route becomes larger. In particular, the difference between the current position and the correct route is considered to be small.

  Note that F2 to F7 are the same as those described above, and optionally included in the current vector.

  FIG. 4 is a functional configuration diagram of a database for calculating a learning vector.

  The learning vectors stored in the learning vector storage units 111, 121 and 131 in the mobile terminal 1 of FIG. 1 are also preferably acquired from other databases. According to the database in FIG. 4, as in FIG. 1, the railway line storage unit 110, the road map storage unit 120, the bus route storage unit 130, the railway traffic learning data storage unit 115, and the road traffic learning data storage unit. 125, a bus traffic learning data storage unit 135, a railway route learning vector storage unit 111, a road map learning vector storage unit 121, and a bus route learning vector storage unit 131. The traffic learning data accumulation units 115, 125, and 135 accumulate a large number of past times and traffic positions accumulated for learning for each moving vehicle.

  As in FIG. 1, the database in FIG. 4 further includes a railway map matching unit 112, a road map map matching unit 122, a bus route map matching unit 132, a train learning vector calculation unit 113, and a car learning vector calculation. Unit 123 and a bus learning vector calculation unit 133. Here, both the current vector calculation unit in FIG. 1 and the learning vector calculation unit in FIG. 4 execute the same processing. As a result, for each moving vehicle, a learning vector including an accumulated error value between the past passage position accumulated for learning and the passage node of the map is output to the learning vector accumulation units 111, 121, and 131.

  The learning vectors stored in the learning vector storage units 111, 121, and 131 are transmitted from the communication interface to the mobile terminal 1 via the network.

Returning to FIG. 1, the posterior probability calculation unit 143 calculates the following posterior probabilities.
The railway (first) posterior probability is calculated on the condition of the railway (first) current vector based on the railway (first) learning vector.
A vehicle posterior probability is calculated on the basis of the vehicle (second) current vector based on the vehicle (second) learning vector.
A bus posterior probability is calculated on the basis of the bus (third) current vector based on the bus (third) learning vector.

  Here, “Posterior probability distribution” means that when there are multiple variables (probability variables) that determine probabilistic values, the values of some of the variables are observed under the observed conditions. The degree to which each value of a variable can take. “Posterior probability” refers to the degree to which each value can be taken under the aforementioned conditions. For example, under the condition where each value of the current vector (values F1 to F7) is observed, the degree of possible values of the moving state can be expressed as the posterior probability distribution of the moving state based on the current vector. Called. The “posterior probability” includes a railway posterior probability based on the current vector, a vehicle posterior probability based on the current vector, and a bus posterior probability based on the current vector.

  The a posteriori probability distribution of the moving state (boarding a moving vehicle) that is based on the current vector is based on the Bayes theorem and the prior probability distribution of the moving state is multiplied by the likelihood of the current vector, and this is normalized (total value or Obtained by setting the integral value to 1. Here, the “priority probability distribution of the moving state” refers to the degree to which each value of the moving state can take under the condition that no value of any random variable is observed. “The likelihood of the current vector” refers to the degree of likelihood that the value of each variable of the current vector, which is a condition for the posterior probability, is observed when the movement state takes a certain value. The multiplication of the prior probability distribution of the moving state and the likelihood of the current vector is referred to as a joint probability distribution of the moving state and the current vector.

  In order to calculate the joint probability distribution of a plurality of random variables, a probability model for expressing the stochastic dependency between the random variables is required. The probability model is composed of a combination of a plurality of random variables and a plurality of conditional probabilities that express a stochastic dependency between the random variables. The probability model is not uniquely determined by the type of the random variable, but varies according to the way of expressing the dependency relationship between the variables. For example, when there are three random variables A, B, and C, “A, B, C” (all variables are independent), “A, B, A → C” (A and B are independent, C is A , "A, A → B, A → C" (A is independent, B and C are dependent on A), "A, B, AB → C" (A and B are independent, C is A and B Can depend on many kinds of dependencies. By defining the probability model, it is possible to derive a formula for calculating a joint probability that is a probability distribution of combinations of values of all variables of a plurality of random variables. For example, in the case of “A, B, A & B → C” (A and B are independent and C depends on A and B), the joint probability distribution P (A, B, C) = P (A) P (B ) P (C | A, B) is calculated. For the probability model, an appropriate expression is selected in advance by learning data (teacher data), and the conditional probability distribution is estimated in advance by the maximum likelihood method or the maximum posterior probability method.

According to the present invention, between the variables of the correct movement state (represented as “Class”), the movement state estimated by the movement state estimation unit (represented as “Predicted”), and the feature vectors of F1 to F7, It is assumed that the dependency relationship “Class, Predicted, Class & Predicted → F1, Class & Predicted → F2,... Class & Predicted → F7” (Class and Predicted are independent and F1 to F7 depend on Class and Predicted) is established. In the case of the aforementioned dependency relationship, the joint probability of the movement state and the current vector is P (Class, Predicted, F1, F2,,,, F7) = P (Class, Predicted) P (F1 | Class, Predicted) P ( F2 | Class, Predicted) ··· P (F7 | Class, Predicted) = P (Class, Predicted) Π i P (F i | Class, a can be calculated by the calculation formula of Predicted). The dependency relationship between the variables is appropriately learned by the maximum likelihood method or the maximum a posteriori probability method according to the learning data, and is not limited to the above-described expression of the dependency relationship.

  FIG. 5 is an explanatory diagram showing calculation of the posterior probability for each moving vehicle.

  According to FIG. 5A, P (class, predicted) is represented. For example, it represents the probability (simultaneous probability) that Class (correct movement state) = automobile and Predicted (estimated movement state) = railroad train will occur at the same time. According to FIG. 5 (a), this probability is 0.03. Further, all the probability values in the table of FIG.

According to FIG. 5B, Π _i P (f _i | class, predicted) is represented. This is the product of the probability of the feature quantity under all conditions of class and predicted. (Here, fi is a feature quantity other than the movement state probability value, and f = {fi} is a vector of feature quantities other than the movement state probability value.) Also, all probability values in the table of FIG. Will add to 1.

  According to FIG.5 (c), the posterior probability distribution which normalized the product of the two tables of Fig.5 (a) and (b) is represented. According to FIG.5 (c), the probability that it is a railroad train is 0.84. Further, all the probability values in the table of FIG.

  The moving vehicle estimation unit 144 compares the first to third posterior probabilities, and estimates that the user who owns the mobile terminal is on the moving vehicle corresponding to the current vector having the highest posterior probability. To do.

  The application unit 145 provides various services to a user who operates the mobile terminal 1 based on the moving vehicle information output from the moving vehicle estimation unit 144.

  Finally, the movement state estimation unit 142 will be described.

  Only when it is estimated by the movement state estimation unit 142 that it is “boarding an electric vehicle”, the processing after the map matching units 112, 122, and 132 is executed.

The movement state estimation unit 142 uses the acceleration data output from the acceleration sensor to estimate which movement state among the n (n ≧ 2) movement states. Examples of the moving state include “stop”, “walking”, “running”, “bicycle boarding”, “train boarding”, “car boarding”, and “bus boarding”. According to the present invention, the movement states of at least “stop”, “electric vehicle boarding”, and “others” are estimated.
"Stop"
"Electric car boarding" = "Train boarding""Automobileboarding""Busboarding"
"Others" = "Walking""Running""Bicycleboarding"
In addition, about a movement state estimation process, it is an existing technique (for example, refer patent document 7).

[Calculation of features]
FIG. 6 is an explanatory diagram for deriving a power spectrum from acceleration data.

First, time-series acceleration data output from the acceleration sensor is divided into a plurality of frames (unit sections). Then, each frame is divided into smaller sub-sections (groups). A power spectrum is calculated from the acceleration data included in each small section by FFT (Fast Fourier Transform). According to this technique, the time series change of the power spectrum of the acceleration data is used to estimate the movement state. By “Fourier transform”, it is possible to extract how many frequency components are included in the acceleration data. When “N = the number of small sections”, the number of calculations is proportional to N ² , but by using “Fast Fourier Transform”, the number of calculations is proportional to N · logN.

[Construction of first probability model of power spectrum using teacher data]
FIG. 7 is an explanatory diagram for deriving a first probability model of a power spectrum.

  Next, a teacher data group to which a correct moving state label is assigned is prepared. The teacher data group includes a power spectrum (feature amount) in a correct movement state.

  Then, a first probability model of the acceleration sensor representing the moving state is constructed. This probability model is constructed by clustering teacher data in all moving states. For clustering, an X-Means method is used. According to the clustering of the K-means method, it is necessary to fix the number of clusters K in advance. On the other hand, the X-Means method is an extension of the K-means method, and can estimate the optimum number of clusters according to data. According to the X-means method, the K-means method is recursively executed with K = 2. The BIC (Bayes information criterion) or AIC (Akaike information criterion) is compared before and after the cluster division, and the division is continued until the value does not improve.

  Here, according to the present method, as an index different from BIC or the like, for each cluster, it is determined whether or not the re-division is to be executed while referring to the deviation of the distribution of the movement state label of the teacher data. Calculate cluster size and number of clusters. After the end of clustering, “average power spectrum” and “label distribution of teacher data” are calculated for each cluster, and set as the first probability model.

  The first probabilistic model described above is insufficient for estimating the movement state with high accuracy and causes deterioration in estimation accuracy. This is because each moving state has an unreliable interval in which it is difficult to distinguish between a confidence interval (frame) in which a power spectrum peculiar to the moving state appears and a zone in another moving state. .

  Therefore, the reliability of the section is evaluated, and the estimation accuracy is improved by using the estimation result of only the section with high reliability (hereinafter referred to as “key frame”). This reliability can be expressed by the probability table described above.

[Construction of second probability model of reliability using teacher data]
FIG. 8 is an explanatory diagram for deriving a second probability model considering the confidence interval.

  According to FIG. 8, the power spectrum and the nearest cluster corresponding to each small section are searched, and the probability table is set as the movement state probability of the small section. This is executed for all the small sections in the frame, and the average value of the moving state probability (hereinafter referred to as “average probability table”) is calculated.

  For example, referring to the average probability table of FIG. 8, “running: 50%” and “walking: 40%” are indicated, and it is difficult to distinguish between “running” and “walking”. However, if this average probability table appears in the “running” state but has characteristics that do not appear in the “walking” state, this average probability table is a reliable “running” state. It can be said that. It is assumed that in the “running” state, movements close to walking are often mixed, but in the “walking” state, movements close to running are rarely mixed.

  Therefore, a second probability model of reliability is constructed using the teacher data described above and the first probability model of the power spectrum. An average probability table for the frame is calculated using the first probability model. This is executed for all frames and all moving states. The average probability table and the set of moving state labels calculated in this way are used as a population, and the average probability table is clustered. Thus, a second probability model of reliability is constructed. Also for this clustering, for each cluster, it is determined whether or not to perform re-division while referring to the deviation of the distribution of the movement state label of the teacher data, and an appropriate cluster size and number of clusters are calculated.

After the end of clustering, an average probability table is calculated for each cluster, and a moving state A having the maximum probability is derived. Moreover, the most frequently appearing movement state B is derived from the movement state label of the teacher data. Then, both movement states are compared.
A = B cluster (average probability table) = reliable cluster A ≠ B cluster (average probability table) = unreliable cluster In this way, a “reliability flag” is assigned to each cluster. Build a reliability probability model.

[Estimation of moving state]
The movement state is estimated using the first probability model and the second probability model described above. As in FIG. 6, the measured acceleration data is divided into frames and small sections, and a power spectrum for each small section is calculated. For each power spectrum, the nearest probability cluster is searched with reference to the first probability model of the power spectrum. The probability table possessed by the nearest neighbor cluster is set as the movement state probability of the small section, and the average value (average probability table) of the movement state probabilities of all the small sections belonging to the frame is obtained.

  FIG. 9 is an explanatory diagram for deriving a confidence interval.

  Next, a cluster having the calculated average probability table and the nearest average probability table is searched with reference to the second probability model of reliability. The reliability is determined with reference to the “reliability flag” of the cluster. If only the confidence interval is referenced and it is in the same movement state, the interval between the two frames is set as the confidence interval (key frame) of the movement state.

  In addition, the movement state estimation part 142 is estimated as 1st movement state (for example, railroad)-> 2nd movement state (for example, car)-> 1st movement state (for example, railroad) according to progress of time. In the case where the second movement state (for example, an automobile) is within a predetermined time, it is also preferable to estimate these movement states as one first movement. Here, the predetermined time may be about 2 minutes, for example.

  As described above in detail, according to the moving vehicle estimation method, the mobile terminal, and the program of the present invention, what kind of moving vehicle is boarded among the electric vehicles on which the user carrying the mobile terminal is boarding It can be estimated. Thereby, the estimation accuracy of the waypoint of the traveling route can be improved.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Mobile terminal 101 Positioning part 102 Wide area communication interface part 103 Acceleration sensor 110 Railway line accumulation part 111 Railway line learning vector accumulation part 112 Railway line map matching part 113 Railway current vector calculation part 120 Road map accumulation part 121 Road map learning vector accumulation part 122 Road Map Map Matching Unit 123 Car Current Vector Calculation Unit 130 Bus Route Accumulation Unit 131 Bus Route Learning Vector Accumulation Unit 132 Bus Route Map Matching Unit 133 Bus Current Vector Calculation Unit 141 Location Accumulation Storage Unit 142 Movement State Estimation Unit 143 A posteriori probability calculation Unit 144 moving vehicle estimation unit 145 application unit 2 GPS satellite 3 base station

Claims (12)

A mobile vehicle estimation method for a mobile terminal having a positioning unit for positioning by receiving positioning radio waves,
The mobile terminal
A first map accumulating unit that accumulates a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating unit that accumulates a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. An accumulation unit;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. A storage unit,
The mobile terminal
A first step of storing the current position measured by the positioning unit according to the passage of time;
A second step of calculating a first error between the current position and a node of the first map, and calculating a second error between the current position and a node of the second map;
A third step of calculating at least a one-dimensional first current vector having the first error accumulation value as an element and calculating at least a one-dimensional second current vector having the second error accumulation value as an element; When,
The first a posteriori probability of boarding the first moving vehicle is calculated on the condition of the first current vector based on the first learning vector, and the second current vector based on the second learning vector is used as a condition. A fourth step of calculating a second posterior probability of boarding the second moving vehicle;
The first posterior probability is compared with the second posterior probability, and it is estimated that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. A mobile vehicle estimation method for a mobile terminal . A mobile vehicle estimation method for a mobile terminal having a positioning unit for positioning by receiving positioning radio waves,
The mobile terminal
A first map accumulating unit that accumulates a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating unit that accumulates a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulating unit that accumulates a first learning vector including a degree of coincidence between a past travel position accumulated for learning and a first map travel route for the first moving vehicle;
A second learning vector accumulating unit that accumulates a second learning vector including a degree of coincidence between a past travel position accumulated for learning and a second map travel route for the second moving vehicle;
The mobile terminal
A first step of storing the current position measured by the positioning unit according to the passage of time;
A second step of calculating a first degree of match between the current position and the route of the first map, and calculating a second degree of match between the current position and the route of the second map;
A third step of calculating a first current vector including a first degree of match and calculating a second current vector including a second degree of match;
The first a posteriori probability of boarding the first moving vehicle is calculated on the condition of the first current vector based on the first learning vector, and the second current vector based on the second learning vector is used as a condition. A fourth step of calculating a second posterior probability of boarding the second moving vehicle;
The first posterior probability is compared with the second posterior probability, and it is estimated that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. A mobile vehicle estimation method for a mobile terminal . The mobile terminal further includes an acceleration sensor,
The pre-stage of the first step further includes a step of estimating a moving state as to whether or not the vehicle is “electric car riding” based on acceleration data output from the acceleration sensor. The moving vehicle estimation method of the mobile terminal according to 2. The estimation of the movement state is a case where it is estimated that the first movement state-> second movement state-> first movement state over time, and the second movement state is within a predetermined time. The mobile vehicle estimation method for a mobile terminal according to claim 3, wherein the mobile state is estimated as one first movement. The first moving vehicle is a road vehicle;
The first map is a road route, with the corners that cross the intersection as nodes,
The second mobile vehicle is a rail vehicle,
The method for estimating a moving vehicle of a mobile terminal according to any one of claims 1 to 4, wherein the second map uses a track as a passing route and a station as a node. For three or more moving vehicles, each moving vehicle has the map storage unit and the learning vector storage unit,
The second step to the fourth step are calculated for each moving vehicle ,
For the fifth step, the posterior probabilities for each moving vehicle are compared, and it is estimated that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having the highest posterior probability. The mobile vehicle estimation method for a mobile terminal according to any one of claims 1 to 5, wherein: The mobile vehicle estimation method for a mobile terminal according to any one of claims 1 to 6, wherein the mobile vehicle includes a railroad train, an automobile, and a bus. The current vector and the learning vector are other elements:
Estimated speed = path length / time required,
Average of the electric field strength between the base station,
Dispersion of electric field strength between base stations,
Average stop interval in movement and / or
The moving vehicle estimation method for a mobile terminal according to any one of claims 1 to 7, further comprising dispersion of stop intervals in movement . A mobile terminal that has a positioning unit that measures a position by receiving positioning radio waves, and that estimates a moving vehicle on which a user who owns the mobile terminal is on board,
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. Storage means;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. Storage means;
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first error between the current position and a node of the first map;
Second map matching means for calculating a second error between the current position and a node of the second map;
First current vector calculation means for calculating at least a one-dimensional first current vector having the first accumulated error value as an element ;
Second current vector calculation means for calculating at least a one-dimensional second current vector having the second error accumulated value as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
A moving vehicle estimation that compares the first posterior probability with the second posterior probability and estimates that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. And a mobile terminal. A mobile terminal that has a positioning unit that measures a position by receiving positioning radio waves, and that estimates a moving vehicle on which a user who owns the mobile terminal is on board,
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulation in which at least a one-dimensional first learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the first map as an element is accumulated for the first moving vehicle. Means,
Second learning vector accumulation in which at least a one-dimensional second learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the second map as an element is accumulated for the second moving vehicle. Means,
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first matching degree between the current position and the route of the first map;
Second map matching means for calculating a second matching degree between the current position and the route of the second map;
First current vector calculation means for calculating a first current vector of at least one dimension having a first degree of match as an element ;
Second current vector calculating means for calculating at least one-dimensional second current vector having the second degree of match as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
A moving vehicle estimation that compares the first posterior probability with the second posterior probability and estimates that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. And a mobile terminal. A program that causes a computer mounted on a terminal having a positioning unit that measures a position by receiving positioning radio waves and an acceleration sensor to function so as to estimate a moving vehicle on which a user having the mobile terminal is on board Because
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector in which at least a one-dimensional first learning vector having an accumulated error value between a past passage position accumulated for learning and a passage node of the first map as an element is accumulated for the first moving vehicle. Storage means;
A second learning vector in which at least a one-dimensional second learning vector having the accumulated error value between the past passage position accumulated for learning and the passage node of the second map as an element is accumulated for the second moving vehicle. Storage means;
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first error between the current position and a node of the first map;
Second map matching means for calculating a second error between the current position and a node of the second map;
First current vector calculation means for calculating at least a one-dimensional first current vector having the first accumulated error value as an element ;
Second current vector calculation means for calculating at least a one-dimensional second current vector having the second error accumulated value as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition Second posterior probability calculating means for calculating a second posterior probability of boarding the second moving vehicle;
A moving vehicle estimation that compares the first posterior probability with the second posterior probability and estimates that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. A program that causes a computer to function as means. A program that causes a computer mounted on a terminal having a positioning unit that measures a position by receiving positioning radio waves and an acceleration sensor to function so as to estimate a moving vehicle on which a user having the mobile terminal is on board Because
First map accumulating means for accumulating a first map including a traffic route and a traffic node in the first moving vehicle;
A second map accumulating means for accumulating a second map including a traffic route and a traffic node in the second moving vehicle;
A first learning vector accumulation in which at least a one-dimensional first learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the first map as an element is accumulated for the first moving vehicle. Means,
Second learning vector accumulation in which at least a one-dimensional second learning vector having a degree of coincidence between a past passage position accumulated for learning and a passage route of the second map as an element is accumulated for the second moving vehicle. Means,
Position cumulative storage means for accumulatively storing the current position measured by the positioning unit as time elapses;
First map matching means for calculating a first matching degree between the current position and the route of the first map;
Second map matching means for calculating a second matching degree between the current position and the route of the second map;
First current vector calculation means for calculating a first current vector of at least one dimension having a first degree of match as an element ;
Second current vector calculating means for calculating at least one-dimensional second current vector having the second degree of match as an element ;
Calculating a first a posteriori probability of boarding the first moving vehicle based on the first current vector based on the first learning vector and using the second current vector based on the second learning vector as a condition A posteriori probability calculating means for calculating a second posteriori probability of boarding the second moving vehicle;
A moving vehicle estimation that compares the first posterior probability with the second posterior probability and estimates that the user possessing the mobile terminal is on the moving vehicle corresponding to the current vector having a high posterior probability. A program that causes a computer to function as means.

Images