JP2015505681A - Apparatus and method for analyzing sensor data - Google Patents

Abstract

An apparatus for detecting an event of interest of interest comprises a first basic event detector (8b) that generates a first basic event stream (10b) using a sensor data stream (4). . The sensor data stream (4) carries raw data at a certain sensor data transfer rate. The raw data relates to one or more properties determined using one or more sensors for a plurality of observation objects (72). The data transfer rate of the first basic event stream (10b) is smaller than the sensor data transfer rate. The apparatus further includes a second basic event detector (8c) that generates a different second basic event stream (10c) using the sensor data stream (4). The data transfer rate of the second basic event stream (10c) is smaller than the sensor data transfer rate. The event detector (16) is operable to determine the event of interest using the first basic event stream (10b) and the second basic event stream (10c).

Description

  Embodiments in accordance with the present invention are generally related to the analysis of sensor data. Specifically, the embodiment according to the present invention relates to an apparatus and method for analyzing sensor data provided at high speed.

  Wireless sensor networks using various technologies are sometimes used to identify locations. For example, a wireless sensor network may be used to locate a person or other object or track a person or other object. Here, “specifying a place” means detecting and determining a geographical place and position. On the other hand, tracking means identifying a place many times at short intervals. Such tracking is performed in order to acquire the position of the moving observation object and other kinematic variables over time. For example, in sports competitions such as soccer, American football, rugby, and tennis, a dedicated location identifying system or a position tracking system may be used to identify the location of a player or other object (for example, a ball).

  Analyzing data gathered on the geographical location and position of players and balls can provide statistical information on overall sporting events such as football matches, and statistical information on individual teams and individual players. The statistical information thus obtained may be of interest for various reasons. For example, there are various commercial interests. This is because a particular statistic and the analysis on that statistic may be particularly relevant for people watching in the stadium and viewers watching in front of home television. Therefore, providing specific statistics may increase your interest in sports competition. In another example, statistical data obtained from raw data regarding a position can be used for training. Specifically, you can analyze your team's movements and opponents, and analyze the performance and health of individual players.

  The above-described location specifying system and position tracking system can be constructed based on various technologies. For example, the location information is determined by analyzing a radio wave signal or a magnetic field. For this reason, there is a method in which a transmitter or receiver is attached to an individual object (for example, a player or a ball) so that the system can determine the location of the individual object ( Usually called a sensor). In addition, there is a method in which a corresponding receiving device or transmitting device is installed at a predetermined place over a geographical area of interest such as a soccer game field. Analysis of signal strength, propagation time, and phase provides a sensor data stream. This sensor data stream indicates the geographic location and location of individual players and objects (players and objects that are detected or observed) at various moments. Note that analyzing signal strength, propagation time, and phase is only part of the various technical techniques available. A data sample for a location may be associated with a time stamp that indicates when the object was at that location or location. In addition to location data, kinematic variables such as speed (speed), acceleration, etc. may be provided as information to be combined as well. The data regarding this place has, for example, an x coordinate, ay coordinate, and a z coordinate. Hereinafter, location data and kinematic data provided by the location sensor system are also referred to herein as raw (sensor) data.

  A specific example of the wireless tracking system will be described. This wireless tracking system may be configured to attach a very small transmitter to a person or an object. This transmitter can be embedded in shoes, uniforms or balls. Signals from this transmitter are sensed by a plurality of antennas. These antennas are arranged so as to surround a place to be observed such as a soccer field. The receiver unit processes the collected signals and determines the arrival time (ToA) of those signals. To track all the objects being observed, the difference in propagation delay is calculated and the position of each transmitter is continuously determined based on this calculation. In one embodiment, the transmitter is designed to operate in the 2.4 GHz frequency band so that it can be used without a license worldwide. In this specific example, CDMA (Code Division Multiple Access) is used as a technique for identifying individual transmitters.

  However, if you track an object with a certain degree of accuracy in a sporting game, a large amount of sensor data is generated immediately. For example, in a soccer game, when tracking a ball with an accuracy of about 3 cm, if the maximum speed of the ball is about 150 km / h, the position of the ball may have to be predicted up to 2000 times per second There is. 150 km / h would be a reasonable maximum speed for a soccer ball. However, in other sports such as tennis and baseball, it is certainly necessary to sample and predict more frequently. Also, in soccer games, it is not enough to track the movement of the ball. It is equally important to track the movement of 22 players. The system described above can basically be used for tracking with an accuracy of 1 cm. For this reason, it may be desirable to use multiple tracking sensors per player. As a result, not only the position of the player can be obtained, but also the posture of the player can be clarified. In this way, for example, it can be determined whether or not the vehicle is offside. If a moderate number of sensors are attached to the players used by both teams and the equipment that must be used in a soccer game, up to 50000 position data and raw data samples may be generated per second. The corresponding sensor data stream sensor data transfer rate can be 50,000 per second.

  In order to analyze the sensor data as described above, it is necessary to grasp all the raw data of a plurality of sensors or all of the sensors at a time each time each determination is made. This is because each time an event is analyzed or a conclusion is made, information based on separate raw data out of the entire raw data of the sensor is required. In other words, when using a conventional analysis framework, all data needs to be delivered simultaneously to all the entities performing the analysis. When the number of elements to be analyzed is increased in order to analyze data from a plurality of viewpoints, it becomes impossible to distribute all the data to these simultaneously. Such a data distribution cannot be performed at the same time as or almost simultaneously with the time when the actual thing is happening. Nonetheless, this may be necessary when pointing to an offside position in a soccer game, or in situations where a conclusion must be drawn instantly during another sport.

  Therefore, there is a need to provide a concept that can more efficiently analyze sensor data streams in order to extract information of interest.

  According to an aspect of the present invention, the above problem can be solved by using an analysis framework for analyzing raw data portions. This raw data portion is transmitted in the sensor data stream at a certain sensor data transfer rate. The analysis framework includes an input layer. The input layer includes at least two or more basic event detection units. The input layer uses only the sensor data stream as an input and generates at least a first basic event stream and a second basic event stream. Furthermore, the analysis framework comprises one or more event layers. Each event layer includes one or more event detection units that detect events of interest (interesting events). Any one of the one or more event detectors may be the only source of the first basic event stream and the second basic event stream, or any other provided by another event detector. Use a stream of events. That is, the analysis framework follows a technique for performing hierarchical analysis based on events. In this way, mutual events are determined or detected in separate hierarchies with different dependent states.

  In general, an event can be defined as an interesting event that occurred instantaneously at a certain point in time. Events are generally associated with a specific amount of change that is perceivable by a sensor. This event may be a basic event obtained from the base of the hierarchy, ie the input layer. In this case, the basic event will be directly based on sensor data (eg, kinematic data provided by the tracking system). As a result, this basic event may indicate a specific amount of change perceived by the sensor. An event of interest (an event of interest) is an event in a higher hierarchy. That is, the event of interest is subordinate to other events or basic events (basic events). More generally, the event of interest is dependent on other events that have been detected previously. Viewed another way, events and basic events contain information obtained by applying a certain criterion to another event or the raw data.

  If an analysis method based on events is used, the original data can be analyzed in parallel while a plurality of processing nodes are communicating on the network. This only requires a high data rate sensor data stream to be provided to the input layer basic event detector. In this input layer, the sensor data stream is processed, and basic events are usually determined at a low data rate. This makes it possible to analyze the original data in parallel. At higher levels, analysis is based on basic events. For this reason, the sensor data stream need not be delivered to all entities that make decisions. There is no need to distribute the sensor data stream to all event detectors. As a result, a larger amount of data related to the position can be analyzed.

  In addition, events are input to the event detection unit in the higher hierarchy at a reduced (relatively low) data transfer rate. As a result, processing can be performed in parallel. For example, in an analysis framework having several CPUs or computation nodes, the event detection unit and the basic event detection unit are distributed to different CPUs and computers in the analysis framework, so that the processing is performed in parallel as described above. Can be done. Because the data transfer rate of the event has been reduced (because it is small), the fact that the event has occurred is transmitted via a conventional network such as a LAN (Local Area Network) based on Ethernet (registered trademark). Even things are possible. In this case, even if it is necessary to analyze at the same time as when the actual thing is happening, it is not necessary to degrade the performance of the analysis framework. As information about the event, only the fact that the criterion of the event detection unit is satisfied (that is, the event has occurred), not the raw data, may be transmitted between the event detection units related to each other. As a result, the amount of data exchanged between the event detectors can be kept low. This is because the length of each message is shorter than the length of one raw data sample. That is, not only can the number of data packets per fixed time be reduced, but also the size of transmitted data packets can be kept small. In other embodiments, the data exchanged for an event may include information about two, three, or more possible (possible) conditions determined by the event detector. That is, the determined event or basic event may include more information than the information of the two components of YES / NO.

  According to another aspect of the present invention, the above-described problem can be solved by an apparatus that detects an event of interest. The apparatus includes a first basic event detection unit that generates a first basic event stream using the sensor data stream. Furthermore, the apparatus includes a second basic event detection unit that generates a stream of different second basic events using the sensor data stream. Here, the data transfer rate of the first basic event stream and the data transfer rate of the second basic event stream are smaller than the sensor data transfer rate.

  The event detection unit is operable to determine the event of interest by using the first basic event stream and the second basic event stream. As described above, the event detection unit can perform determination and analysis based on a basic event that occurs at a low data transfer rate. The event detector does not need to directly access the high data rate sensor data stream.

  Here, the term “data transfer rate” should be interpreted as meaning not only the data transfer rate at a certain moment but also the average data transfer rate. The average data transfer rate means an average of instantaneous data transfer rates over a certain period. Depending on the set of basic events selected, a specific basic event may be determined by the transfer rate of sensor data in a short period of time. However, the average data transfer rate for that particular basic event during the period of observation, eg, a soccer game, may be much smaller.

  According to some aspects of the present invention, the basic event detector can also use only kinematic data as raw data. This is for immediately providing the analysis result of the tracking data in a sporting game or the like. Kinematic data can be defined as any collection of position information, velocity information, and acceleration information of an observation object (observed object). This data may be provided in a one-dimensional, two-dimensional, or three-dimensional space. As a result, the kinematic data used here is a vector or a scalar, depending on the configuration used. According to another aspect of the present invention, the basic event detection unit may derive speed information and acceleration information by itself based on the provided position information. In this case, the sensor may be a simple tracking device or a positioning device.

  In the hierarchical analysis, basic events are accurately selected in order to obtain basic information necessary for all decisions made in a higher dimension (higher hierarchy). An analysis is performed based on these basic events.

  According to an aspect of the present invention, the fact that there is a second observation object in a predetermined neighborhood area of the first observation object is that the position information of the first observation object and the second observation object When the position information indicates, a proximity event (an event of being near) is generated as a basic event. The predetermined neighborhood area depends on what kind of event or sport game is being observed (that is, the predetermined neighborhood area changes depending on the content of the event or sport game. May be a complex solid or may be defined by the distance between the first and second observation objects). Proximity may be used as input data when detecting many other events. For example, proximity events are used as input data in determining whether a player, batter, or other equipment has hit a ball or other observed object.

  According to another aspect of the present invention, the fact that the second observation object is outside a predetermined neighboring area of the first observation object is that the position information of the first observation object and the second When the position information of the observed object indicates, a remote event (an event of being far away) is generated as a basic event.

  According to another aspect of the present invention, when the change in the velocity (direction or size) of the observation object or the change in acceleration exceeds a predetermined threshold, the direction change event (the direction is changed) Event) is generated. This criterion is generally useful for sports and tracking applications. However, the threshold size varies depending on the observed sport. Therefore, the threshold value may vary depending on the sport, and the threshold value may vary to reflect environmental changes such as raining in one game.

  According to another aspect of the present invention, when the speed of the observed object is less than a predetermined threshold, idling is used as a basic event to indicate that the object is inactive or stationary. An event (an event that a person is free) is generated. In some embodiments, an idling event is generated only if the speed of the observed object is maintained at least below the predetermined threshold for a predetermined minimum duration.

  According to another aspect of the present invention, when the speed of the observation object becomes larger than a predetermined threshold, an operation activation event (an event that the moving state is entered) is generated as a basic event. In some embodiments, the motion activation event is generated only if the speed of the observed object remains at least greater than the predetermined threshold for a predetermined minimum duration.

  According to another aspect of the present invention, when at least one component of the position information of the observation object exceeds a predetermined threshold value or becomes lower than the predetermined threshold value, the basic event is used as a line over event (line Event) is generated.

Next, some embodiments according to the apparatus and method will be described as examples. At that time, reference is made to the following drawings.
FIG. 1 is a schematic diagram showing an analysis framework for analyzing a sensor data stream. FIG. 2 is a block diagram schematically illustrating one aspect of a method for detecting an event of interest using a sensor data stream. FIG. 3 is a diagram schematically illustrating an example of a scheme for detecting an event. In this example, an event is detected in which a shot is shot toward a goal in a soccer game. FIG. 4 is a diagram schematically illustrating an example of a scheme for detecting an event. In this example, an event that a double pass was made in a soccer game is detected. FIG. 5 is a diagram schematically showing a plurality of events. According to FIG. 5, it is possible to efficiently determine that the observation object is in a predetermined area. FIG. 6 is a diagram schematically illustrating an aspect of the event detection unit.

  FIG. 1 is a schematic diagram illustrating an analysis framework 2 that analyzes raw data portions to identify (identify) at least one event of interest (an event of interest). This raw data portion is transmitted in the sensor data stream 4 at a certain sensor data transfer rate. This raw data portion is data relating to one or more properties of an object (observed object) observed using one or more sensors. In this embodiment and the embodiments described later, it is assumed that only raw kinematic variables for players and balls observed in a soccer game are included in the raw data.

  Of the framework 2, the input layer 6 includes basic event detectors 8a-8e. They generate a stream of basic events with only the sensor data stream 4 as input (input source). In other words, only the basic event detection units 8a-8e in the input layer of the analysis framework 2 receive the raw data input at a high sensor data transfer rate.

  The first basic event detection unit 8b and the second basic event detection unit 8c receive only the sensor data stream 4 as an input (input source), the stream 10b related to the first basic event, and the second basic event detection unit 8c. A stream 10c related to the event is generated. Any technique suitable for providing data can be used to provide the sensor data stream 4. For example, it is possible to use an Ethernet network using copper cables or fiber cables. Various other wired and wireless technologies can also be used. These technologies use standardized communication protocols or specially set communication protocols.

  The analysis framework 2 includes two event layers 12 and 14. Each event layer includes an event detection unit that detects an event of interest. Each event detection unit of the event layers 12 and 14 uses a stream provided by the basic event detection unit or a stream provided by another event detection unit of the event layers 12 and 14 as an input (input source). . For example, the event detection unit 16 uses the stream 10b of the first basic event detection unit 8b and the stream 10c of the second basic event detection unit 8c to determine the event of interest. The stream 10 b is further provided to the event detection unit 18. Of course, the components may be interdependent in any other way. However, for convenience of explanation, only some interdependent forms and connection forms are given here as examples.

  The event detection units may be connected by wired Ethernet. Moreover, you may connect by the other network connection based on the arbitrary techniques known in the technical field which concerns on this application. Individual event detection units may operate as separate software processes on different computation nodes. Several event detection units may operate on the same computing node. Furthermore, the event detection unit may be configured as dedicated hardware, hardware that can be controlled by a program, or the like. Therefore, the interdependency between the event detection units shown in FIG. 1 should be understood as an example of physical hardware connection or an example of logical connection. These indicate the flow of information from one event detection unit to another event detection unit and how the event detection units are interdependent.

  FIG. 2 schematically illustrates one aspect of a method for detecting an event of interest using the sensor data stream 4. This sensor data stream 4 carries raw data at a certain sensor data transfer rate. Raw data relates to one or more properties of an object (observed object) observed using one or more sensors.

  In the first basic event generation step 20, a stream 22 relating to the first basic event is generated. Here, the data transfer rate of the stream 22 relating to the first basic event is smaller than the sensor data transfer rate. In the second basic event generation step 24, a stream 26 relating to the second basic event is generated. Here, the data transfer rate of the stream 26 relating to the second basic event is smaller than the sensor data transfer rate.

  In the event generation step 28, the event of interest 30 is determined using the stream 22 related to the first basic event and the stream 26 related to the second basic event. Either the first basic event generation step 20 or the second basic event generation step 24 may be executed first. That is, the data relating to the specific basic events of the first and second streams used to determine the event of interest 30 need not be generated at the same time and need not be distributed at the same time.

  For this purpose, the event detection unit can be configured as a state machine (state machine) as shown in FIG. 6, for example.

  FIGS. 3 to 5 described below show some aspects of the basic events that have been chosen appropriately. These basic events are used to obtain events of interest at a higher level for soccer matches.

  3 and 4 show examples of analysis frameworks. In these examples, in order to identify (identify) a large number of events of interest, position information of players and balls in a soccer game is analyzed. The embodiment shown in FIGS. 3 and 4 does not show all the elements (entities) for making judgments, interdependencies and connection states between a plurality of event detectors in individual hierarchies. . FIG. 3 and FIG. 4 are mainly drawn to explain the hierarchical analysis method more clearly. Specifically, FIG. 3 and FIG. 4 are diagrams for easily showing that, when data is analyzed using some basic events, some complex information can be obtained from some basic events.

  FIG. 3 shows an example of an analysis framework for analyzing position information of players and balls in order to identify an event (interest of interest) that a shot to a goal is blocked in a soccer game. In the input layer of the framework, a “direction change event” 40 is detected for the ball. This concludes a change in the direction of the ball, for example, by comparing the provided acceleration to a threshold or by calculating the first and second derivatives of the ball position data. To do. Further, a line crossing event 42 is generated or obtained. This event 42 indicates that the ball has crossed a predetermined virtual line in the playing field.

  The line crossing event 42 identifies (identifies) that the ball is in a specific area, thereby generating an area event 44. How to identify (identify) a specific area will be described in more detail with reference to FIG. Since the ball is in the playing field, it can be determined that the observed ball is the moving ball 46. This moving ball 46 is a ball that is actually used during a soccer game and is not a spare ball stored outside the playing field. If the approach event 48 occurs with respect to the player and the ball, and further to the ball and the goal, it is determined that the change in the direction of the ball indicated in the direction change event is caused by a specific player kicking the ball. Event 50 corresponds to a specific player kicking the ball. Similarly, it is determined by the event 52 that a shot has been released toward the goal. After that, if an approach event occurs for the ball and goalkeeper, it can be determined that even if the player kicks the ball, the shot is blocked and no points are scored. As is apparent from FIG. 3, the complex information that is finally obtained is gathered from only a few basic events that have been chosen appropriately.

  FIG. 4 shows an example of an analysis framework for analyzing position information of players and balls in order to identify (identify) a double pass by two players in a soccer game. The determination (detection) that a specific player has kicked the ball is executed as described with reference to FIG. Further, the event 56 in which the height of the ball reaches the peak is determined from the change in the sign of the ball speed or the change in the component of the ball speed vector. This can indicate that the ball has been shot rather than just dribbling. The proximity event 48 determines (identifies) an event 58 that the ball has been held and has ended. This event 58 indicates that the ball is near a particular player. The fact that a specific player is controlling the ball is represented by reference numeral 60, and information that the acceleration of the ball changes while the same one player holds the ball (the ball has various accelerations). Information). The pass 62 can be determined from the information that the ball is moving from when one player stops holding the ball until another player starts holding the ball. This is detected by a height event 56. Finally, the double pass 64 can be determined from the fact that the same players subsequently passed each other twice. In the example of FIG. 4, complex information is analyzed and determined, but the complex information is also analyzed and determined based on only a few basic events that have been properly selected and revealed.

  FIG. 5 shows that the information that the ball 72 is located in a specific area 74 in the playing field can be estimated (collected) very efficiently from just one basic event. The underlying basic event is a line crossing event. This event indicates that the ball has crossed a straight line virtually projected to divide the playing field into multiple areas. Region 72 is formed by four lines 76-82. These lines 76-82 intersect at right angles to form a rectangular region 74. When using a Cartesian coordinate system as shown in FIG. 5, this is (information on whether or not the ball is located within a specific area), whether the component of the ball position vector is greater than a predetermined threshold. It can be actually obtained by judging whether it is small. However, in the embodiment described here, there may be various other shapes as the shape of the observed region. Then, when determining whether or not the line has been exceeded, there is a possibility that, for example, setting of a value used for determination (determinate) or the like must be performed by a different calculation method.

  To determine whether the ball 72 is in the area 74 using conventional techniques, it is repeatedly determined for all areas in the playing field where the ball is located relative to the four corners of each area. There is a need. That is, in the conventional method, for example, in order to determine (conclude) that the ball 72 is in the region 74 when all the values used for the determination have the same sign, each of the four corners of the region 74 is determined. It was necessary to calculate a value to be used for judgment. It would be almost impossible to do such a process immediately, for example for 2000 positions with a ball per second.

  However, if the embodiment according to the present invention is used, it is possible to determine the position of the ball based on the basic event “line over”. Whether or not the basic event “line over” has occurred is determined for each line dividing the field into a plurality of areas. Examples of those lines are lines 76-80 that delimit the region 74. For each line, when the ball crosses the line, an over-line event or a relative position event is determined (detected). As shown in FIG. 5, the ball moves from the position 84 where the ball was previously to the current position in region 74. As a result, the output state of the basic event detection unit related to the line 82 changes. That is, the basic event detection unit related to the line 82 generates a line crossing event.

  When the basic event detection unit related to the line 82 generates an element that exceeds the line, the event generation unit receives this event (an event that exceeds the line). , 80, the event determined by the basic event detection unit is also received. As a result, the event generation unit can generate an event that the ball 72 is currently in the region 74. The event detection unit can be configured as a state machine (state machine), for example. This point will be described in more detail when described with reference to FIG. Compared with the conventional method, there is a possibility that the calculation load when detecting where the ball is located with respect to a predetermined area is greatly reduced. An example of the predetermined area is a penalty area.

That is, the conventional approach requires the following steps for each defined region:
For each straight line (line) in this area:
Calculate the determinate value for the position of the observed object;
The calculated values used for judgment are compared, and it is determined (concludes) whether the object is within the region.

On the other hand, in the embodiment according to the present invention, as illustrated in FIG. 5, for example, k lines that divide the competition field into n areas can be defined. Then you can perform the detection process in the following steps:
Activate k basic event detectors. Each detection unit has, as a basic event, to determine whether the ball is on one side (first side) or the other side (second side) of the line. Related to one line.
Activate n event detectors. This is for determining an event indicating that the ball has entered an area related to (in charge of) each event detection unit. Each event detection unit is connected to a basic event detection unit for a boundary line of a region to which the event detection unit relates (in charge).
All the connected basic event detection units, when providing the corresponding information, determine the event together with the event detection unit.

  The event detection unit 100 may be configured as a state machine (state machine), and FIG. 6 schematically illustrates an example thereof. This state machine depends on four input events, that is, input values A, B, C, and D. These input values may be provided from one or a plurality of basic event detection units, for example. When this state machine enters state 102, the state machine, ie, the event detection unit, generates or determines (determines) an event of interest related thereto. This is naturally determined depending on the input of the event detection unit. This state machine is initialized to a starting configuration 104. When event A occurs, the state machine transitions to state 106. When event B occurs, the state machine transitions from state 106 to state 108. When event C occurs, the state machine transitions to the final state 102. The state machine then determines (determines) what the associated event is. When event D occurs, the state machine is re-initialized from states 106 and 108 to the initial activation settings 104. This event D may be interpreted as a reset signal. In the particular embodiment shown in FIG. 6, in order to arrive at the final state 102, the order in which the signals A-C occur is important. However, it is possible to use another state machine that is irrelevant regardless of the order in which the events are input. In general, if a state machine is used as an event detection unit, it may be incorporated into hardware. By doing so, there is a possibility that the calculation load in the analysis framework is reduced and the overall processing speed is increased.

  Although only some of the events of interest have been described, any further events of interest can be determined based on the base event. Optional further events of interest include, for example, a single pass from one player to another player, a double pass, a shot to a goal, a dribbling, a free kick, a kick-off, and a corner kick. Can be mentioned.

  In this specification, a soccer game has been described as a representative example, but other embodiments according to the present invention can of course be applied to any other sports game. For example, it can be applied to basketball, handball and volleyball games, rugby, American football, ice hockey, field hockey, polo, baseball, tennis and golf.

  The description and drawings merely show examples of the principles according to the present invention. Therefore, it is possible for those skilled in the art to create various combinations that are suitable for the principle and spirit of the present invention and are included in the scope of the present invention, even if not explicitly shown in these explanations and drawings. Will. Further, all examples appearing in the above description and drawings are introduced solely for educational purposes. In other words, these examples are provided so that it is easy for the reader to understand the principles of the present invention and what concepts those skilled in the art have created to contribute to the technical field. . The present invention should not be construed as limited to the specific examples and conditions set forth. Further, all the explanations introducing the principle of the present invention, one aspect and embodiment of the present invention, and specific examples thereof are written with the intention of referring to the equivalents of the above.

  Each functional block expressed as “means for... Fulfilling a specific function” is understood as a functional block comprising an electrical circuit configured to perform a specific function. To do. Therefore, the expression “means for doing something” can be understood as “a means suitable for something or operable for something”. Thus, means for performing a specific function do not necessarily perform that function (at a specific moment).

  Functions of various components such as functional blocks depicted in the drawings can be provided by using dedicated hardware. Examples of dedicated hardware include a processor and hardware capable of executing software in conjunction with software suitable for the purpose. Where functions are provided by a processor, the functions may be provided by one dedicated processor, one shared processor, or multiple individual processors. Some of the plurality of individual processors may be shared processors. Further, the use of the terms “processor” and “control unit” should not be construed to mean only hardware capable of executing software. The terms “processor” and “control unit” may include, for example, the following meanings: digital signal processor (DSP) hardware, network processor, application specific integrated circuit ( ASIC), field programmable gate array (FPGA), read only memory (ROM) for storing software, random access memory (RAM), non-volatile storage. Other conventionally used hardware and specially set hardware may also be included in the meanings of the terms “processor” and “control unit”.

  It should be understood by those skilled in the art that the block diagrams introduced in the present application conceptualize an electric circuit that is a specific example of the embodiment of the principle according to the present invention. Similarly, flowcharts, process flow diagrams, state transition diagrams, pseudo code, and the like show examples of various processes. These processes may be substantially represented in a computer readable medium, and even if a computer or processor is not introduced in this application, the process is executed on a computer or processor.

  Further, the claims to be described later are incorporated into the detailed description. Each claim can be an independent embodiment. On the other hand, there are dependent claims that refer to combinations with one or more other claims, but other embodiments similarly include combinations of the dependent claims with the contents of each of the other dependent claims. There is a case. Such combinations are assumed to be presented herein. However, when the intention of removing a specific combination is clearly indicated, the specific combination is not included in the content of the present application. Furthermore, the features described in the claims are described with the intention of being included in any other independent claims. Even if the above claim does not designate a particular independent claim as a dependent party, the independent claim shall include the features of the above claim. In particular, when a dependent claim refers to an encoder or transmitter, the corresponding features of the associated encoder or transmitter as well as part of the disclosure of the specification are intended to be included.

  The method disclosed in the specification and the claims can be executed by an apparatus provided with means for performing each step of the method.

  Further, even if a plurality of steps and functions are disclosed in the specification and the claims, they should not necessarily be performed in a specific order. Therefore, even if a plurality of steps and functions are disclosed, there is no limitation that these steps should be performed in a specific order. However, this excludes cases where it is impossible to interchange processes and functions for technical reasons.

  Furthermore, in some embodiments, a single step may include multiple sub-steps. In some cases, one process can be decomposed into a plurality of sub-processes. Unless stated otherwise, substeps are also intended to be included as part of a single disclosed step.

Claims (26)

An apparatus for detecting an event of interest (30; 102), which is an event of interest,
The apparatus comprises a first basic event detector (8b) that generates a first basic event stream (10b) using the sensor data stream (4), the sensor data stream (4) Carries raw data at a certain sensor data transfer rate, the raw data being one or more properties determined using one or more sensors for a plurality of observation objects (72). The data transfer rate of the first basic event stream (10b) is smaller than the sensor data transfer rate,
The apparatus includes a second basic event detection unit (8c) that uses the sensor data stream (4) to generate a different second basic event stream (10c). The data transfer rate of the stream (10c) is smaller than the sensor data transfer rate,
Further, the apparatus is operable to determine the event of interest (30; 102) using the first base event stream (10b) and the second base event stream (10c). An apparatus comprising a detection unit (16; 100).   The first basic event detector (8b) and the second basic event detector (8c) are operable to use only kinematic data as the raw data. The apparatus according to 1.   The apparatus according to claim 2, wherein the kinematic data includes position information, velocity information, and acceleration information of an observation object.   The first basic event detector (8b) and the second basic event detector (8c) are operable to use only the sensor data stream (4) as an input. The apparatus according to any one of claims 1 to 3.   When the position information of the first observation object and the position information of the second observation object indicate that there is a second observation object in a predetermined neighboring region of the first observation object, the first observation object The apparatus according to any one of claims 1 to 4, wherein a proximity event is generated as a basic event or the second basic event.   6. The apparatus of claim 5, wherein the predetermined neighborhood region is defined by a predetermined maximum distance between the first observation object and the second observation object.   When the position information of the first observation object and the position information of the second observation object indicate that the second observation object is outside a predetermined neighboring area of the first observation object, the first observation object The apparatus according to any one of claims 1 to 6, wherein a remote event is generated as the basic event or the second basic event.   The direction change event is generated as the first basic event or the second basic event when the change in the speed of the observation object or the change in the acceleration exceeds a predetermined threshold value. The apparatus of any one of these.   9. The idling event is generated as the first basic event or the second basic event when the speed of the observation object is smaller than a predetermined threshold value. 9. The device described.   10. The apparatus of claim 9, wherein the idling event is generated only when the velocity of the observation object is at least less than the predetermined threshold for a predetermined minimum duration.   The operation activation event is generated as the first basic event or the second basic event when the speed of the observation object is larger than a predetermined threshold value. The device according to item.   12. The apparatus of claim 11, wherein the motion activation event is generated only when the velocity of the observation object is greater than at least the predetermined threshold for a predetermined minimum duration.   When the position information component of the first observation object or the position information component of the second observation object exceeds or falls below a predetermined threshold, the first basic event or the second observation event The apparatus according to claim 1, wherein a line crossing event is generated as a basic event.   The first basic event detector (8b) detects a proximity event for the first object and the second object, and the second basic event detector (8c) for the second object. When a direction change event is detected, a hit event is detected as the event of interest (30; 102), and the hit event indicates that the first object has hit the second object. The apparatus according to any one of claims 1 to 13.   The additional event detection part (18) which detects an additional interested event (30; 102) using the first basic event and the interested event, further comprising: The apparatus according to item 1.   When a proximity event is detected by the first basic event detection unit (8b) and a hit event is detected by the event detection unit (16), an object holding event is used as the additional interest event (30; 102). The device of claim 15, wherein the device is generated.   When a remote event is detected by the first basic event detection unit (8b) and a hit event is detected by the event detection unit (16), an object loss event is detected as the additional event of interest (30; 102). The device of claim 15, wherein the device is generated.   Device according to any one of the preceding claims, characterized in that the event detector (16; 100) is configured as a state machine (100). A method for detecting an event of interest (30; 102), which is an event of interest, comprising:
The method includes generating a first elementary event stream (10b; 20) using a sensor data stream (4), wherein the sensor data stream (4) is a sensor data transfer. Carrying raw data at a speed, wherein the raw data is related to one or more properties determined using one or more sensors for a plurality of observation objects; The data transfer rate of the basic event stream (10b; 20) is smaller than the sensor data transfer rate,
The method includes generating a different second base event stream (10c; 26) using the sensor data stream (4), wherein the second base event stream (10c; 26) The data transfer rate is smaller than the sensor data transfer rate,
The method further comprises determining the event of interest (30; 102) using the first basic event stream and the second basic event stream. An analysis framework for analyzing a raw data portion to identify at least one interesting event of interest (30; 102), the raw data portion comprising a sensor data stream (4) Are transmitted at a certain sensor data rate, and each raw data portion is related to one or more properties determined using one or more sensors for the observed object. ,
The analysis framework comprises an input layer (6) comprising at least two or more basic event detectors (8a-8e), which inputs only the sensor data stream (4). To generate at least a first basic event stream (10b) and a second basic event stream (10c),
Furthermore, the analysis framework comprises one or more event layers (12, 14) comprising one or more event detectors (16, 18) for detecting the event of interest (30; 102), Any one of the two or more event detectors (16, 18) has as input only the first basic event stream (10b) and the second basic event stream (10c), or some other event. An analysis framework characterized by using a stream of other events provided by the detection unit (16).   The one or more event detection units of the one or more event layers (12, 14) are connected to the at least two basic event detection units (8b, 8c) of the input layer (6) via a network. The analysis framework according to claim 20, wherein the analysis framework is connected to   The analysis framework of claim 21, wherein the network is a wired network.   23. An analysis framework according to claim 21 or claim 22, wherein the network communication is based on an internet protocol.   24. The data transfer rate of the first basic event stream (10b) and the second basic event stream (10c) is less than the sensor data transfer rate. The analysis framework according to claim 1.   The first basic event detection unit (8b) and the second basic event detection unit (8c) are operable to use only kinematic data of an observation object as the raw data. The analysis framework according to any one of claims 20 to 24.   A computer program for performing the method of claim 19 when executed on a computer.

Images