JP2009163537A - Fall determination system, fall determination method, and fall determination program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To highly precisely detect fall of a person regardless of a fall operation pattern or the like. <P>SOLUTION: This fall determination system is provided with a sensor mounted on the trunk of a person for outputting a value corresponding to the perpendicular displacement quantity of the trunk; a candidate selection means for selecting an output value satisfying first conditions from among the output values of the sensor for the first time as fall candidate data which can be related with the operation at falling; a candidate displacement quantity calculation means for calculating the perpendicular candidate displacement quantity of the trunk corresponding to each selected fall candidate data by using the output value of the sensor; a fall displacement quantity calculation means for calculating the fall displacement quantity by using only the candidate displacement quantity satisfying the second conditions among the calculated candidate displacement quantity; and a fall decision means for deciding the fall of the person based on the calculated fall displacement quantity. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a fall determination system, a fall determination method, and a fall determination program for determining a person's fall.

  In recent years, with the increase in the number of elderly people, home care, entry to welfare facilities for the elderly, elderly people living alone, and elderly households are increasing according to the circumstances of the elderly. In any case, it is difficult for family members or caregivers to always attend. Accordingly, there is a need for a system in which when an elderly person falls due to a sudden seizure or the like when an attendant or the like is not on the side, a family member or a caregiver or the like detects this and promptly reports it. An example of such a system is disclosed in Patent Document 1 below, which detects a fall using a device attached to the trunk of an elderly person or the like who may fall and notifies a family member or caregiver or the like. Yes.

  In the system disclosed in Patent Document 1 below, a device attached to the trunk of an elderly person measures vibration associated with the movement of the elderly person. Then, when this measured value becomes equal to or higher than the first threshold value, it is determined that there is a suspicion that the elderly has fallen, and immediately after that, a measured value equal to or lower than the second threshold value lower than the first threshold value is predetermined. When the time continues, it is determined that the elderly person has fallen and an alarm signal is transmitted to the caregiver.

  However, since the system of Patent Document 1 below detects only vibration as an index, there is a disadvantage that it cannot be determined whether the vibration is caused by a fall or vibration caused by another movement (such as lying down). In addition, when an elderly person moves on the floor or the like immediately after falling, the apparatus measures vibration according to the movement. At this time, there is a possibility that the device measures vibration exceeding the second threshold before a predetermined time has elapsed after the suspected fall, and does not detect that the elderly has fallen.

Further, for example, the following Patent Documents 2 and 3 disclose a system for determining a fall with higher accuracy than the system of Patent Document 1 below by using an index other than vibration. In these systems, a device equipped with a sensor that detects accelerations in three axial directions orthogonal to each other and angular velocities around the three axes is attached to the trunk of an elderly person. Then, when the acceleration detected by the acceleration sensor exceeds the first threshold, it is determined that an event suspected of falling has occurred. Next, the vertical displacement of the trunk of the elderly (hereinafter referred to as “vertical displacement”) is calculated using the acceleration and the output value of the angular velocity sensor at that time, and the calculated vertical displacement is the second If it is equal to or greater than the threshold value, the elderly person's posture has collapsed so much that it cannot be normal, and it is determined that the elderly person has fallen and is notified to the caregiver.
Japanese Patent Laid-Open No. 9-305875 JP 2006-227752 A JP 2005-237576 A

  The systems disclosed in Patent Documents 2 and 3 can determine relatively accurately for a simple fall (for example, a case of falling to the floor without tripping from a walking state). However, there are individual differences in the movements of people at the time of a fall, and there are various fall motion patterns depending on the situation at the time of the fall. is there. Therefore, in the systems disclosed in Patent Documents 2 and 3, there is a possibility that even if an elderly person actually falls, the fall of the elderly person cannot be detected depending on the falling action pattern.

  For example, it is assumed that an elderly person falls to the floor after a movement that triggers a fall (for example, after a stumbling). In this case, if the acceleration at the time of the trip detected by the sensor is equal to or greater than a predetermined threshold, the vertical displacement amount is calculated using the acceleration. However, since the vertical displacement calculated at this time is likely to be a low value (a value lower than the second threshold value) corresponding to the displacement at the time of the trip, the device should not detect that the elderly has fallen. There is. In addition, the amount of vertical displacement from a low posture after falling over to falling to the floor or the like is also likely to be below the second threshold, and the device may not detect that an elderly person has fallen, as described above. . Even when a fall is detected, there is an inconvenience that the tripping time and the falling time are actually different. For this reason, provision of a more accurate fall determination system is desired.

  Accordingly, in view of the above circumstances, the present invention provides a fall determination system, a fall determination method, and a fall determination program capable of detecting a fall of an elderly person with high accuracy regardless of a fall motion pattern or the like. It is an issue.

  A fall determination system according to an aspect of the present invention that solves the above-described problem is a system for determining a person's fall, and is a value that is attached to a human trunk and corresponds to a vertical displacement amount of the trunk. And a candidate selection means for selecting an output value satisfying the first condition from the sensor output values for the first time as fall candidate data that may be related to the action at the time of the fall. Candidate displacement amount calculating means for calculating the vertical candidate displacement amount of the trunk corresponding to each fall candidate data using the output value of the sensor, and a second condition among the calculated candidate displacement amounts A system comprising: a fall displacement amount calculating means for calculating a fall displacement amount using only a candidate displacement amount to be satisfied; and a fall determination means for judging a person's fall based on the calculated fall displacement amount It is.

  According to the fall determination system configured as described above, the candidate displacement amount corresponding to the output data that can be related to the action at the time of the fall is obtained, and the actual action at the time of the fall is expressed from the obtained candidate displacement amount. It is possible to calculate the fall displacement amount using only the candidate displacement amount having a high possibility, and to perform the fall determination based on the calculated fall displacement amount. That is, at the stage of obtaining the final displacement amount, candidates that are highly likely not to be related to the operation at the time of the fall can be favorably excluded. For this reason, it is possible to perform the fall determination of various fall motion patterns with high accuracy from a simple fall motion pattern to a complicated fall motion pattern.

  Here, it is desirable that the fall determination system further includes a fall suspect determination unit that determines whether or not an event suspected of falling has occurred based on the output of the sensor. In this case, the candidate selection means selects the fall candidate data for the output value for the first time after the occurrence of the event is detected by the fall suspect determination means.

  The fall determination system further includes a correction value calculation unit that estimates an error in the output value of the sensor and calculates a correction value, and a correction unit that corrects the fall displacement amount based on the calculated correction value. It is also desirable.

  The correction value calculation means may be configured to calculate a correction value based on an output value for a second predetermined time immediately before the occurrence of an event suspected of falling.

  The correction unit may include an offset value calculation unit that calculates an offset value for correcting the amount of overturning displacement based on the correction value calculated by the correction value calculation unit.

  Moreover, in the fall determination system, the first condition is, for example, to have an output value equal to or higher than the first threshold value.

  The candidate displacement amount calculation means is configured to, for example, divide the output values for the first time into fall candidate data and calculate the candidate displacement amount using the output value group of each section.

  The candidate displacement amount calculation means is configured to calculate a candidate displacement amount corresponding to each fall candidate data using, for example, output value groups for a predetermined time before and after each fall candidate data.

  Further, the candidate displacement amount calculating means uses each output candidate group between output values between two output values that are close to the front and back of each fall candidate data and are equal to or lower than the second threshold and lower than the first threshold. A candidate displacement amount corresponding to the data is calculated.

  Further, the candidate displacement amount calculating means determines whether or not the plurality of output value groups include output values in a period overlapping each other, and determines that the plurality of output value groups include output values in an overlapping period. Values can be merged into a single output value group.

  Here, the second condition in the above fall determination system is, for example, to have a value equal to or greater than a third threshold value.

  Further, the fall displacement amount calculating means may be configured to calculate the fall displacement amount by adding only candidate displacement amounts that satisfy the second condition.

  The sensor includes, for example, an acceleration sensor or jerk sensor. In this case, the fall suspect determination means monitors the output of the acceleration sensor or jerk sensor, determines that an event suspected of falling has occurred when an output value equal to or greater than a predetermined threshold is detected, and the candidate displacement amount calculation means The candidate displacement is calculated by integrating the output value of the acceleration sensor or jerk sensor with time.

  The sensor further includes, for example, an angular velocity sensor. In this case, the candidate displacement amount calculation means uses the output value of the angular velocity sensor to convert the vector of the output value of the acceleration sensor or jerk sensor so as to correspond to a predetermined static coordinate system, and the converted acceleration sensor or The candidate displacement amount is calculated by integrating the output value of the jerk sensor with time.

  The fall determination system uses a correction selection unit that selects whether or not to correct the fall displacement amount calculated by the fall displacement amount calculation unit, and a predetermined correction value when selected to correct the fall displacement amount. It is also possible to further include a correcting means for correcting the amount of overturning displacement.

  In addition, it is desirable that the fall determination system further includes a fall notification means for notifying the fact when the fall determination means determines that the person has fallen.

  Here, the fall notification means includes at least one of, for example, an alert display means that can display an alert, an alert sound output means that outputs an alert sound, and an alert message transmission means that transmits an alert message to a predetermined contact, When the fall determination means determines that a person has fallen, at least the alert display by the alert display means, the output of the alert sound by the alert sound output means, the transmission of the alert message to the predetermined contact by the alert message transmission means, It is good also as a structure which performs one.

  A fall determination method according to an aspect of the present invention that solves the above-described problem is a method for determining a person's fall, and a sensor attached to the trunk of the person is displaced in the vertical direction of the trunk. A sensor output step for outputting a value according to the amount and an output value satisfying the first condition from the sensor output values for the first time are selected as fall candidate data that can be related to the action at the fall. A candidate selection step, a candidate displacement amount calculating step for calculating the vertical candidate displacement amount of the trunk corresponding to each selected fall candidate data, using the output value of the sensor, and the calculated candidate displacement amount A fall displacement amount calculating step for calculating the fall displacement amount using only the candidate displacement amount satisfying the second condition, and a fall determination step for determining a person's fall based on the calculated fall displacement amount. Is the method.

  According to such a fall determination method, the candidate displacement amount corresponding to the output data that can be related to the motion at the time of the fall is obtained, and the actual motion at the time of the fall can be expressed from the obtained candidate displacement amounts. The fall displacement amount is calculated using only the high candidate displacement amount, and the fall determination is performed based on the calculated fall displacement amount. That is, when the fall determination method of the present invention is applied, candidates that are highly likely not to be related to the operation at the fall can be favorably excluded at the stage of obtaining the final displacement amount. For this reason, it is possible to perform the fall determination of various fall motion patterns with high accuracy from a simple fall motion pattern to a complicated fall motion pattern.

  Here, it is desirable that the fall determination method further includes a fall suspect determination step for determining whether or not an event suspected of falling has occurred based on the output of the sensor. In this case, in the candidate selection step, the fall candidate data is selected for the output value for the first time after the occurrence of the event is detected in the fall suspect determination step.

  The fall determination method further includes a correction value calculation step for estimating an error in the output value of the sensor to calculate a correction value, and a correction step for correcting the fall displacement amount based on the calculated correction value. It is desirable that

  In the correction value calculation step, a correction value may be calculated based on an output value for a second predetermined time immediately before the occurrence of an event suspected of falling.

  In the correction step, an offset value for correcting the fall displacement amount may be calculated based on the correction value calculated in the correction value calculation step.

  In the fall determination method, the first condition is, for example, to have an output value equal to or greater than a first threshold value.

  In the candidate displacement amount calculating step, the output value for the first time may be divided for each fall candidate data, and the candidate displacement amount may be calculated using the output value group of each section.

  In the candidate displacement amount calculating step, a candidate displacement amount corresponding to each fall candidate data may be calculated using an output value group for a predetermined time before and after each fall candidate data.

  Further, in the candidate displacement amount calculating step, each fall candidate data is output by using an output value group between two output values that are close to each other before and after each fall candidate data and are equal to or lower than a second threshold value that is lower than the first threshold value. It is also possible to calculate a candidate displacement amount corresponding to.

  Further, in the candidate displacement amount calculating step, it is determined whether or not the plurality of output value groups include output values in the overlapping period, and when it is determined that the output values in the overlapping period are included, the plurality of outputs The value group may be merged into a single output value group.

  In the fall determination method, the second condition is to have a value equal to or greater than a third threshold, for example.

  In the fall displacement amount calculation step, the fall displacement amount may be calculated by adding only candidate displacement amounts that satisfy the second condition.

  Further, the sensor is an acceleration sensor or jerk sensor, and in the suspected fall determination step, the output of the acceleration sensor or jerk sensor is monitored, and an event suspected of falling is detected when an output value equal to or greater than a predetermined threshold is detected. It may be determined that it has occurred, and in the candidate displacement amount calculation step, the candidate displacement amount may be calculated by time-integrating the output value of the acceleration sensor or jerk sensor.

  The sensor also includes an angular velocity sensor. In the candidate displacement amount calculating step, the output value of the acceleration sensor or jerk sensor is associated with a predetermined static coordinate system using the output value of the angular velocity sensor. The candidate displacement amount may be calculated by integrating the output values of the converted acceleration sensor or jerk sensor after time.

  Further, the fall determination method includes a correction selection step for selecting whether or not to correct the fall displacement amount calculated in the fall displacement amount calculation step, and a predetermined correction value when selected to correct the fall displacement amount. It is good also as a method further including the correction | amendment step which correct | amends the said amount of fall displacement using.

  In addition, the fall determination method is preferably a method further including a fall notification step of notifying that when it is determined in the fall determination step that the person has fallen.

  Here, in the fall notification step, when it is determined in the fall determination step that a person has fallen, an alert is displayed on the display, an alert sound is output from a speaker, and a predetermined contact is alerted via the network. At least one of sending a message may be executed.

  In addition, a fall determination program according to an aspect of the present invention that solves the above-described problem is a program for causing a computer to execute the fall determination method.

  According to such a fall determination program, the candidate displacement amount corresponding to the output data that can be related to the motion at the time of the fall is obtained, and the actual motion at the time of the fall can be expressed from the obtained candidate displacement amounts. The fall displacement amount is calculated using only the high candidate displacement amount, and the fall determination is performed based on the calculated fall displacement amount. That is, when the fall determination program of the present invention is applied, candidates that are highly likely not to be related to the action at the fall can be favorably excluded at the stage of obtaining the final displacement amount. For this reason, it is possible to perform the fall determination of various fall motion patterns with high accuracy from a simple fall motion pattern to a complicated fall motion pattern.

  According to the fall determination system, the fall determination method, and the fall determination program of the present invention, the candidate displacement amount corresponding to the output data that can be related to the operation at the time of the fall is obtained, and the actual displacement amount is obtained from the obtained candidate displacement amount. The fall displacement amount is calculated using only the candidate displacement amount that is highly likely to represent the action at the time of the fall, and the fall determination is performed based on the calculated fall displacement amount. That is, when the fall determination program of the present invention is applied, candidates that are highly likely not to be related to the action at the fall can be favorably excluded at the stage of obtaining the final displacement amount. For this reason, it is possible to perform the fall determination of various fall motion patterns with high accuracy from a simple fall motion pattern to a complicated fall motion pattern.

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

  FIG. 1 is a block diagram showing a configuration of a fall determination system 1 according to an embodiment of the present invention. The fall determination system 1 includes a data collection device 100 and a fall determination device 200 as shown in FIG. The data collection device 100 collects data for determining the fall of the elderly and transmits it to the fall determination device 200. Then, the fall determination device 200 makes a fall determination using the received data, and when it is determined that the fall has occurred, transmits the notification to a family member or a caregiver (nurse or the like). Hereinafter, the configuration and operation of the data collection device 100 and the fall determination device 200 will be described in detail.

  The data collection device 100 is used in a state where the data collection device 100 is attached to the trunk of an elderly person by an attachment portion (not shown). The data collection device 100 includes a CPU (Central Processing Unit) 103 that performs overall control of the entire device.

  Each component is connected to the CPU 103 via the bus 121. The CPU 103 implements various functions by performing data communication via the bus 121 and exchanging with each component. These components include a RAM (Random Access Memory) 105, an acceleration sensor 107, an angular velocity sensor 109, an RTC (Real Time Clock) 111, a ROM (Read Only Memory) 113, an A / D conversion unit 115, and a data transmission / reception unit 117. There is. The bus 121 includes a power supply line in addition to the data line. The power supply unit 119 supplies power to each component through a power supply line. The RTC 111 is driven by battery backup, and always keeps time even when the power is turned off.

  The acceleration sensor 107 is a type including a piezoelectric element, for example. The piezoelectric element is distorted by an amount corresponding to the acceleration applied to the acceleration sensor 107 itself, and the voltage generated by the piezoelectric effect at this time is an analog signal representing acceleration (hereinafter, abbreviated as “acceleration analog signal”). Similarly, this is abbreviated). The acceleration sensor 107 includes a piezoelectric element corresponding to directions of three axes (three axes of x, y, and z) orthogonal to each other, and outputs an acceleration analog signal in three directions.

  The angular velocity sensor 109 is also of a type including, for example, a piezoelectric element, and outputs a voltage generated according to the angular velocity of the angular velocity sensor 109 itself as an angular velocity analog signal. The angular velocity sensor 109 includes a piezoelectric element that is coaxial with the three axes (roll angle φ, pitch angle θ, yaw angle ψ), and outputs an angular velocity analog signal around the three axes.

  Note that the three-dimensional orthogonal coordinate system of the acceleration sensor 107 changes according to the attachment state and the movement of the elderly (more precisely, the direction of the acceleration sensor 107 itself). For this reason, it does not necessarily coincide with a static three-dimensional orthogonal coordinate system having the gravity direction as the Z axis. Further, in this specification, the three-dimensional orthogonal coordinate system of the acceleration sensor 107 is expressed as (x, y, z), and the static three-dimensional orthogonal coordinate system is expressed as (X, Y, Z) to clearly distinguish them. Here, “static” is a term given for convenience in order to distinguish it from a three-dimensional orthogonal coordinate system of the acceleration sensor 107, that is, a coordinate system that moves in accordance with the change in the direction of the acceleration sensor 107 itself.

  Further, the acceleration sensor 107 and the angular velocity sensor 109 are relatively fixed by the housing of the data collection device 100. Therefore, the relative positional relationship does not depend on the mounting state of the data collection device 100 or the movement of the elderly person, and the roll angle φ, pitch angle θ, and yaw angle Ψ are always around the same axis as the x, y, and z axes. It becomes the angle of. In the following description, parameters with lowercase letters (x, y, z) correspond to the three-dimensional orthogonal coordinate system of the acceleration sensor 107, and parameters with uppercase letters (X, Y, Z) are static 3 The parameters correspond to the dimensional orthogonal coordinate system.

  When the power is turned on, the data collection device 100 always collects data for determining the fall of an elderly person. Here, FIG. 2 shows a flowchart of data collection processing executed by the data collection device 100 in order to collect the data.

  As shown in FIG. 2, the acceleration sensor 107 and the angular velocity sensor 109 respectively generate an acceleration analog signal and an angular velocity analog signal corresponding to the movement of the elderly person, and output them to the A / D converter 115 (step 1 and thereafter). In the description and drawings, the step is abbreviated as “S”). Next, the A / D converter 115 samples the input acceleration analog signal and angular velocity analog signal at a predetermined sampling period, quantizes them, converts them into acceleration digital signals and angular velocity digital signals, and passes them to the CPU 103 (S2).

  When the CPU 103 receives the acceleration digital signal and the angular velocity digital signal, the CPU 103 acquires time data counted by the RTC 111 (or constantly acquires the time data). Then, the received acceleration digital signal, angular velocity digital signal, and a set of time data (hereinafter referred to as “determination data set”) when these digital signals are received are written into the RAM 105 (S3).

  Since the above series of operations (that is, acquisition of acceleration analog signals, A / D conversion, creation and writing of determination data sets) is always executed when the power is turned on, information on acceleration and angular velocity is stored in the RAM 105 according to predetermined sampling. It will be written in a cycle. However, since the capacity of the RAM 105 is limited, the determination data set cannot be continuously written. In order to write new data, it is necessary to periodically delete unnecessary data.

  In the present embodiment, the CPU 103 performs memory management. For example, when im determination data sets are stored in the RAM 105, a predetermined number of determination data sets are deleted. Specifically, the CPU 103 refers to the time data of each determination data set, and preferentially erases a predetermined number of determination data sets determined to be old among the im determination data sets. As a result, the storage area of several blocks or several segments is released, and a new determination data set can be written.

  Once the number of determination data sets stored reaches im, the storage area release and data writing are repeatedly executed. Therefore, the RAM 105 is the latest determination data set, and always holds the determination data set for at least several seconds. The number of samplings of the determination data set corresponding to several seconds exceeds the minimum number required to execute the fall determination process of FIG. Since the RAM 105 is a volatile memory, its contents are lost when the power is turned off.

  Next, a suspected fall detection process for detecting occurrence of a suspicious event as a fall will be described. The fall suspect detection process is a process in which the CPU 103 develops and executes a fall suspect detection program stored in the ROM 113 in the work area of the RAM 105.

  FIG. 3 shows a flowchart of the suspected fall detection process. Note that, for example, a dedicated IC independent of the CPU 103 may be configured to execute the suspected fall detection process of FIG. 3 alone or in cooperation with the CPU 103.

  The fall suspect detection process is executed in parallel with the data collection process of FIG. 2 only when the power of the fall determination device 200 is turned on (that is, the fall suspect detection program is in a resident state). As shown in FIG. 3, when the power is turned on, the CPU 103 monitors the absolute value of the acceleration digital signal value from the A / D conversion unit 115 (hereinafter referred to as “acceleration absolute value”) (S5), and the acceleration absolute value. Is determined to be greater than or equal to the threshold value q (S6).

  When the CPU 103 determines that the absolute acceleration value is equal to or greater than the threshold value q (S6: YES), a large acceleration that is rarely detected in normal operation (such as walking) is detected, and an event that is suspected of falling has occurred. The process proceeds to S7. Then, in the process of S7, for determination for several seconds (for example, 5 seconds before and after) around the determination data set corresponding to the acceleration absolute value determined to be equal to or greater than the threshold value q (hereinafter referred to as “falling suspect data set”). A data set (hereinafter referred to as “determination data set group”) is transmitted from the data transmitting / receiving unit 117 to the fall determination device 200.

  On the other hand, if the CPU 103 determines that the acceleration absolute value is lower than the threshold value q (S6: NO), it returns to the process of S5 because there is no significant change in the movement of the elderly person and there is no possibility of falling. Continue monitoring the absolute acceleration.

  For example, the data transmission / reception unit 117 modulates a carrier with a determination data set group and transmits the modulation wirelessly (or by wire). When the data collection device 100 and the fall determination device 200 are configured to be capable of wireless communication, there is a merit that the devices are wireless and do not limit the action range of the elderly.

  A description will be added about the processing of S7. In the process of S <b> 7, the CPU 103 reads out the determination data set before the suspected fall data set from the RAM 105 and transmits it to the fall determination device 200. On the other hand, there is no determination data set after the suspected fall data set when the suspected fall data set is stored in the RAM 105. With respect to these determination data sets, the CPU 103 sequentially transmits each time a new determination data set is stored, or transmits it collectively when a predetermined number of data sets are stored.

  Note that the CPU 103 does not return to the process of S5 until all determination data sets to be transmitted in the process of S7 are transmitted. Therefore, even if the acceleration digital signal of the determination data set for at least 5 seconds after the suspected fall data set includes a threshold q or higher, the CPU 103 transmits another fall suspect data set group corresponding thereto. There is nothing.

  Here, FIG. 4A shows a graph for more specifically explaining the suspected fall detection process of FIG. The vertical axis of the graph represents the absolute acceleration value acquired by the data collection device 100, and the horizontal axis represents time.

  According to the example of FIG. 4A, before the time tu, the acceleration absolute value is less than the threshold value q, and the motion of the elderly is regarded as a normal motion (walking or the like). At time tu, the acceleration absolute value becomes equal to or greater than the threshold value q, and the CPU 103 detects that an suspicious event has occurred in the elderly for the first time at that time. Then, for example, determination data sets (determination data set group) for 5 seconds before and after the time tu are transmitted to the fall determination device 200. There are some acceleration absolute values that are greater than or equal to the threshold value q after the time tu, but any value is included within 5 seconds from the time tu. Therefore, according to the example of FIG. 4A, the CPU 103 transmits only the determination data set group caused by the acceleration absolute value at the time tu to the fall determination device 200 and transmits another determination data set group. There is nothing.

  Next, the fall determination device 200 will be described in detail. The fall determination device 200 is a terminal that functions as a host of the data collection device 100, and the main functions are the fall determination of elderly people and notification to family members or caregivers.

  As shown in FIG. 1, the fall determination device 200 includes a CPU 203 that performs overall control of the entire device. Each component is connected to the CPU 203 via the bus 221. The CPU 203 implements various functions by performing data communication via the bus 221 and exchanging with each component. These components include a RAM 205, a ROM 207, an RTC 211, a display 213, a communication unit 215, a data transmission / reception unit 217, and a speaker 223. Note that the bus 221 includes a power supply line in addition to the data line, similarly to the bus 121 described above. The power supply unit 219 supplies power to each component through a power supply line. As with the RTC 111, the RTC 211 is also driven by battery backup.

  Similarly to the ROM 113 described above, the ROM 207 stores a program that the CPU 203 develops and executes in the work area of the RAM 205. The programs stored in the ROM 207 include a fall determination program 209a, a vertical displacement calculation program 209b, an offset calculation program 209c, a notification determination program 209d, a candidate selection program 209e, and an integration range determination program 209f. These programs are resident in the RAM 205 at the same time as the power of the fall determination device 200 is turned on, and execute the fall determination process shown in the flowchart of FIG. Hereinafter, the overturn determination process of FIG. 5 will be described. In the present embodiment, the main routine of the fall determination process is executed by the fall determination program 209a, and each subroutine is executed by other programs.

  In this specification, a plurality of embodiments will be described with respect to the overturn determination process of FIG. First, a first embodiment of the present invention (hereinafter referred to as “present example 1”, and other embodiments are also described in the same manner) will be described.

  As shown in FIG. 5, the CPU 203 is in a reception standby state of the determination data set group (S11). The transmission data of the data collection device 100 transmitted in the process of S7 in FIG. 3 is received by the data transmission / reception unit 217 and carrier demodulated. The data transmitting / receiving unit 217 passes the determination data set group obtained by carrier demodulation to the CPU 203. When the CPU 203 receives the determination data set group (S11: YES), the CPU 203 transfers it to the RAM 205 and performs writing (S12). When the data transfer and writing of all the determination data set groups are completed, the process proceeds to S13. Here, the following description will be given assuming that the number of determination data sets included in the determination data set group is im.

  The process of S13 is a part of the fall determination process converted into a subroutine, and is executed by the candidate selection program 209e. Note that the process of S13 (and the process executed by the vertical displacement calculation program 209b, the offset calculation program 209c, the notification determination program 209d, and the integration range determination program 209f) does not have to be a subroutine, and falls. It may be a process executed in the main routine of the determination process. Further, the fall determination program 209a, the vertical displacement calculation program 209b, the offset calculation program 209c, the notification determination program 209d, the candidate selection program 209e, and the integration range determination program 209f are not separate program modules as in this embodiment. For example, it may be configured as a single or more detailed program module.

  FIG. 6 shows a subroutine of S13 of FIG. The subroutine shown in FIG. 6 selects a time at which the elderly person thinks that a relatively large movement is performed as a parameter for determining a fall.

  The CPU 203 refers to each determination data set written in the RAM 205 to calculate the time candidate. For this purpose, first, initial values of the respective parameters are set (S21). Specifically, the CPU 203 sets the count value i of the reference number counter and the count value n of the candidate number counter to “0” and “1”, respectively. Also, “im” is set as the maximum value i_max of the count value i. These counters are built-in counters of the CPU 203, for example. The reference number counter counts the number of determination data sets referred to in the subroutine of S13. The candidate number counter counts the number of candidates selected at the above time.

  After the initial value setting, the CPU 203 increments the count value i by 1 (S22), and executes the determination process of S23 using each determination data set in the determination data set group. Here, the CPU 203 manages each determination data set in the determination data set group with serial numbers “1” to “im”. A serial number corresponding to the creation order is assigned to each determination data set.

  In the process of S23, the CPU 203 refers to the acceleration absolute value Ai of the determination data set managed by the serial number that matches the count value i, and determines whether the value is equal to or greater than the candidate determination threshold k (S23). . For example, when the count value i is “1”, the CPU 203 executes the process of S23 using the oldest determination data set in the determination data set group. When the acceleration absolute value Ai is less than the candidate determination threshold k (S23: NO), the CPU 203 returns to the process of S22, increments the count value i by 1, and uses the next determination data set to perform S23. Execute the process.

  When the acceleration absolute value Ai is equal to or greater than the candidate determination threshold value k in the processing of S23 (S23: YES), the CPU 203 performs a relatively large operation at the time corresponding to the current determination target determination data set. Judge. Then, the time data included in the data set is recorded in the RAM 205, for example, as candidate time data tkn (S24).

  After recording the candidate time data tkn, the CPU 203 next determines whether or not the count value i is “im” (S25). When the CPU 203 determines that the count value i is “im” (S25: YES), the CPU 203 determines that the subroutine of FIG. 6 has been executed with reference to all the determination data sets written in the RAM 205, and FIG. Return to the main routine. When the CPU 203 determines that the count value i is not “im” (S25: NO), the CPU 203 determines that there is an unreferenced determination data set in the subroutine of FIG. 6, and returns to the process of S22 to return to the subroutine. repeat.

  Note that the CPU 203 executes error processing when none of the candidate time data tkn is obtained in the subroutine of FIG. Specifically, the overturn determination process in FIG. 5 is once reset (for example, the determination data set group is deleted), and the process returns to S11 in FIG. That is, the CPU 203 determines that the elderly person has not fallen, and returns to the data reception standby state.

  FIG. 7 shows a diagram for specifically explaining the subroutine of FIG. The upper graph in FIG. 7 is the same graph as FIG. 4A, and points a, b, and c are attached as explanatory symbols. Each point indicates a region having an acceleration absolute value Ai equal to or greater than the candidate determination threshold k in the graph. Further, the lower graph in FIG. 7 is a graph obtained by extracting the absolute acceleration value Ai around each point. In the lower graph of FIG. 7, for convenience of explanation, the acceleration absolute value Ai is quantized with a lower number of bits than other graphs, and is sampled at a longer period.

  According to the example of the graph in the lower part of FIG. 7, in the subroutine of FIG. 6, the CPU 203 has a total of four candidate times: candidate time data tk1 at point a, candidate time data tk2 at point b, candidate time data tk3 at point c, and 4. Get the data. Hereinafter, for convenience of explanation, the times indicated by the candidate time data tk1, tk2, tk3, and tk4 are referred to as candidate times tk1, tk2, tk3, and tk4, respectively.

  Further, as is apparent from the upper graph of FIG. 7, before the time tu, the acceleration absolute value Ai is below the threshold value q and cannot be greater than or equal to the candidate determination threshold value k. In other words, the acceleration absolute value Ai cannot be greater than or equal to the candidate determination threshold k before an event suspected of falling. Therefore, in another embodiment, the subroutine of FIG. 6 may be executed only for the determination data set after the suspected fall data set. In this case, effects such as resource saving and processing speed improvement are expected.

  FIG. 8 shows a subroutine of S14 of FIG. The subroutine of FIG. 8 is executed by the vertical displacement calculation program 209b. The subroutine in FIG. 8 is a displacement amount calculation process for calculating the vertical displacement amount Z of the trunk of the elderly as a parameter for determining a fall, and returns the calculated vertical displacement amount Z to the main routine as a return value.

  In order to calculate the vertical displacement amount Z, the CPU 203 performs integration processing using the determination data set group. As a parameter for that, first, an integration range is set as shown in FIG. 8 (S31).

  FIG. 9 shows a subroutine of S31 of FIG. The subroutine of FIG. 9 is executed by the integration range determination program 209f. The subroutine of FIG. 9 sets the integration range used in the subroutine of FIG. 8, and returns the set value to the subroutine of FIG.

  As shown in FIG. 9, the CPU 203 performs initial value setting (S51). Specifically, the count value i of the built-in counter is set to “1”. Further, the number n of candidate time data acquired by the above-described processing is set as the maximum value n_max of the count value i. In addition, it demonstrates along the example of FIG. 7 here. Accordingly, the CPU 203 sets “4” as the maximum value n_max.

  Following the processing of S51, the CPU 203 creates an array index of integration start time and integration end time, and sets the table value of the index to “1” (S52). Then, the entry of the integration start time tdi and the integration end time tdi 'corresponding to the candidate time tki is input to the array index corresponding to the current table value (S53). In the first embodiment, the candidate time tk (i−1) is input as the integration start time tdi, and the candidate time tki is input as the integration end time tdi ′.

  Further, the CPU 203 treats the time when the suspicious event has occurred (that is, the time tu) as the candidate time tk0. Therefore, for example, when the count value i is “1”, the respective entries of the integration start time tdi and the integration end time tdi ′ are the candidate times tk0 and tk1.

  Following the processing of S53, the CPU 203 determines whether or not the count value i has reached “4”, that is, the maximum value n_max (S54). If the count value i has reached “4” (S54: YES), it is determined that the subroutine of FIG. 9 has been executed for all candidate times, and the process returns to the subroutine of FIG. If the count value i is less than “4” (S54: NO), it is determined in the subroutine of FIG. 9 that there is a candidate time in which the entry of the array index is not input, and the process of S55 is executed. In the process of S55, the CPU 203 increments the count value i and the table value by 1, returns to the process of S53, and repeats the subroutine.

With the subroutine of FIG. 9, four sets of integration start time and integration end time corresponding to the candidate times tk1, tk2, tk3, and tk4 are obtained. Specifically, the following times are obtained as entries in each table. In each table, the former is the integration start time tdi and the latter is the integration end time tdi ′.
[Table 1] ... Candidate times tk0, tk1
[Table 2] ... Candidate times tk1, tk2
[Table 3] ... Candidate times tk2, tk3
[Table 4]... Candidate times tk3 and tk4

  When the CPU 203 returns from the subroutine of FIG. 9 to the subroutine of FIG. 8, the integration range is set, and the values n and t of the counter (for example, the built-in counter of the CPU 203) are set to “1” and “tk (n−1)”, respectively. (S32, 33). Immediately after the return from the subroutine of FIG. 9, the integration range is set using the values in Table 1. That is, the CPU 203 sets the integration start time and integration end time to tk0 and tk1, respectively, and sets the count values n and t to “1” and “tk0”, respectively.

After completing the setting of the integration range and the count value, the CPU 203 then obtains the vertical displacement amount Z using the following equations (1) and (2). First, the process of S34 is executed to obtain a coordinate conversion coefficient. The coordinate conversion coefficient is a static value obtained by coordinate-converting each vector of acceleration digital signals (acceleration components in the x-, y-, and z-axis directions with “Ax”, “Ay”, and “Az” attached to each). It is a coefficient for making it correspond to a three-dimensional orthogonal coordinate system. In order to obtain a coordinate conversion coefficient, the CPU 203 first integrates each component of the angular velocity digital signal (angular velocity around the x, y, and z axes) to obtain a roll angle φ, a pitch angle θ, and a yaw angle ψ. Next, the obtained roll angle φ, pitch angle θ, and yaw angle ψ are given to the rotation transformation matrices R _φ , R _θ , R _ψ , and the coordinate calculation coefficients Rot _t {φ _t , θ _t , ψ _t } Is calculated (see equation (1)).

After calculating the coordinate conversion coefficient, the CPU 203 then converts the acceleration digital signal vector {Ax (t), Ay (t), Az (t)} obtained at time t into the coordinate calculation coefficient Rot _t {φ _t , θ _t , Coordinate transformation is performed using Ψ _t } to obtain vectors {AX (t), AY (t), AZ (t)} corresponding to the static three-dimensional orthogonal coordinate system (S35). As a result, the vertical acceleration AZ (t) of the trunk of the elderly at time t corresponding to the Z axis is obtained (see formula (2)).

  A specific description will be added to the processing of S34 and S35. For example, when the count value t of the counter is tk0, in the process of S34, the CPU 203 reads out the angular velocity digital signal corresponding to the time tk0 from the determination data set group, gives the value to the equation (1), and coordinates Get the conversion factor. Next, in the process of S35, an acceleration digital signal corresponding to time tk0 is read out, and the obtained coordinate conversion coefficient is given to equation (2) to obtain vertical acceleration AZ (t).

  The CPU 203 integrates the vertical acceleration AZ (t) obtained in the process of S35 by the second order over time. Thereby, the vertical displacement amount Δdn (here, Δd1) of the trunk of the elderly who is displaced at the time Δt starting from the time tk0 is obtained (S36). Each calculated vertical displacement amount Δd1 is held in the RAM 205, for example. Note that Δt is a value corresponding to the sampling period of the determination data set.

  The CPU 203 adds the calculated vertical displacement amount Δd1 to the vertical displacement amount d1 (S37), and proceeds to the process of S38. Next, when the current count value t is tkn (here, tk1), it is determined that the vertical displacement amount d1 is obtained by calculating all the vertical displacement amounts Δd1 in the integration range (S38: YES), and the vertical displacement amount. d1 is entered into Table 1. Next, the process proceeds to S39.

  In the process of S38, if the count value t is not tk1, the CPU 203 determines that the integration is not completed (S38: NO), counts up the count value t (S41), and returns to the process of S34. Repeat the subroutine. The value to be counted up is a value corresponding to Δt. Therefore, the determination data set referred to in the processes of S34 and S35 moves to the next determination data set every time it is counted up.

  In the process of S39, the CPU 203 searches the array index and determines whether there is a table in which the vertical displacement amount dn is not input. If there is an uninput table (S39: YES), the count value n is incremented by 1 (S40), the process returns to S33 and the subroutine is repeated. If the vertical displacement amount dn has been input in all tables (S39: NO), the CPU 203 determines that all vertical displacement amounts dn have been calculated by repeating the subroutine, and proceeds to the processing of S42. According to the example here, the CPU 203 calculates four displacement amounts d1, d2, d3, and d4.

  Next, the CPU 203 determines whether or not each of the calculated vertical displacement amounts has a value equal to or greater than a threshold value d_th ′ (S42). When a vertical displacement amount equal to or greater than the threshold value d_th ′ is detected, the detected vertical displacement amount is added to obtain a vertical displacement amount Z (S43). The CPU 203 returns the calculated vertical displacement amount Z to the main routine of FIG. If any vertical displacement amount falls below the threshold value d_th ′ in the processing of S42, the CPU 203 executes error processing in the same manner as in the subroutine of FIG.

  Here, for example, it is assumed that the vertical displacement amounts d3 and d4 have values greater than or equal to the threshold value d_th ′. In this case, the vertical displacement amounts d3 and d4 are added in the process of S43, and the vertical displacement amount Z is obtained. In other words, the vertical displacement amounts d1 and d2 are less than the threshold value d_th ′. This means that the operation corresponding to the vertical displacement amounts d1 and d2 is an operation irrelevant to the fall of the elderly.

  For example, when an elderly person stumbles or hits an object and receives an impact, the trunk of the elderly person instantaneously moves greatly, so the acceleration sensor 107 outputs a relatively large value. Based on this output, the fall determination device 200 detects the vertical displacement amounts d1, d2, and the like. However, the actual amount of displacement of the trunk itself may be small depending on a trip or an impact. In this case, there is a high possibility that the elderly will return to a state in which his / her posture is corrected and can normally operate. That is, a relatively small stumbling or impact is likely to be an operation that is not included in a series of actions at the time of the fall and is irrelevant to the fall.

  Therefore, in the first embodiment, relatively small displacement amounts such as the vertical displacement amounts d1 and d2 are regarded as irrelevant to the overturning operation, and are not added when calculating the vertical displacement amount Z in the process of S43. The vertical displacement amounts added in the process of S43 are only relatively large vertical displacement amounts d3 and d4 that are likely to be included in a series of actions at the time of the fall. The vertical displacement amounts d3 and d4 are likely to be relatively large trips and impacts that are associated with a fall, or the fall itself.

  When the CPU 203 returns from the subroutine of FIG. 8 to the main routine of FIG. 5, the CPU 203 next executes an offset value calculation process of S15. This offset value calculation process is also a subroutine similar to the vertical displacement calculation program 209b, and is executed by the offset calculation program 209c.

  FIG. 10 shows a subroutine of the offset value calculation process. The subroutine of FIG. 10 calculates the offset value O with respect to the vertical displacement amount Z, and returns the calculated offset value O to the main routine as a return value. The reason for calculating the offset value O is to perform a more accurate fall determination in consideration of an error of the acceleration sensor 107 or the like that occurs due to a change in the surrounding environment or the operating environment, for example.

  More specifically, the subroutine of FIG. 10 reduces the error of the acceleration sensor 107 and the like due to changes in the surrounding environment and the operating environment before and after the process of S35 in FIG. This process is executed to improve the accuracy of the fall determination (in other words, reduce the probability of erroneous determination that the elderly person has fallen in spite of the fact that the elderly person has not actually fallen in the process of S17 described later).

  As shown in FIG. 10, the CPU 203 acquires the vertical acceleration AZ (t) corresponding to each sampling time from time ts to tu (S61). The time ts is, for example, a time that is back by a predetermined time from tu. The method for obtaining the vertical acceleration AZ (t) is the same as the processing of S31 to S35 in FIG. Next, the CPU 203 obtains an average value and standard deviation of each vertical acceleration AZ (t) acquired in the process of S61, and then obtains a correction value of the vertical acceleration AZ (t) based on the average value and standard deviation ( In other words, an error in the output value of the acceleration sensor 107 or the like is estimated). Subsequently, the obtained correction value of the vertical acceleration AZ (t) is second-order integrated to obtain an offset value O for correcting the vertical displacement amount Z (S62). The CPU 203 returns the acquired offset value O to the main routine of FIG. 5 as a return value.

  When the CPU 203 returns from the subroutine of FIG. 10 to the main routine of FIG. 5, the CPU 203 next executes the process of S16 to obtain the vertical displacement amount Z ′. The vertical displacement amount Z ′ is obtained by subtracting the offset value O from the difference between the return values of the subroutines, that is, the vertical displacement amount Z. In other words, it can be said that the process of S16 artificially changes a threshold value d_th (described later) using the offset value O. For this reason, the threshold value d_th may be corrected using the offset value O as an alternative to the vertical displacement amount Z in the process of S16.

  After acquiring the vertical displacement amount Z ′, the main routine proceeds to notification determination processing in S <b> 17. This notification determination process is also made into a subroutine similar to the vertical displacement calculation program 209b and the like, and is executed by the notification determination program 209d.

  FIG. 11 shows a subroutine of the notification determination process. In the subroutine of FIG. 11, the CPU 203 determines whether the elderly person has fallen, and notifies the family or caregiver that the person has fallen according to the determination result. Specifically, the CPU 203 determines whether or not the vertical displacement amount Z ′ obtained in S16 is equal to or greater than the threshold value d_th (S71). When the vertical displacement amount Z ′ is equal to or greater than the threshold value d_th (S71: YES), the CPU 203 determines that the elderly person has fallen because the vertical displacement amount of the trunk of the elderly person is too large to be normal.

  The CPU 203 displays an alert on the display 213 to notify the family or caregiver of the fall of the elderly (S72), and controls the communication unit 215 to notify via the network (S73). Next, the execution of the CPU 203 returns to the main routine of FIG. 5 (more precisely, the process of S11), and shifts to the reception standby state of the determination data set group.

  On the other hand, when the vertical displacement amount Z ′ is less than the threshold value d_th (S71: NO), the CPU 203 determines that the vertical displacement amount of the trunk of the elderly is in the normal range and has not fallen, and enters the main routine. Return to the reception standby state of the determination data set group.

  FIG. 4B is a graph showing the relationship between the vertical displacement amount Z ′ and time. According to the example of FIG. 4B, the vertical displacement amount Z ′ exceeds the threshold value d_th. Accordingly, in this case, the processing of S72 and S73 is executed, an alert is displayed on the display 213, and the elderly or the fall of the elderly is notified to the family or caregiver.

  The network includes various communication networks including, for example, a carrier communication network, an intranet, the Internet, and the like. The ROM 207 (or a non-volatile memory not shown) stores, for example, a contact information such as a family member or a caregiver (for example, an address of a mobile phone or PHS (Personal Handyphone System), a telephone number, an e-mail address, etc.). ing. The CPU 203 reads out the contact information stored in the ROM 207 or the nonvolatile memory, and activates, for example, a mailer to create and transmit a predetermined alert message.

  In addition, the alert display on the display 213 indicates that, for example, a contact information such as a family member or a caregiver (contact information different from the contact information stored in the non-volatile memory) is desired to be contacted immediately. A message for informing surrounding people is included. When a nearby person (for example, a friend or the like) notices that an elderly person falls, he or she can see the contact information displayed on the display 213 and contact the contact information. Such personal contact can be an insurance in the event that a failure occurs in the network used by the communication unit 215 and the alert message does not reach the family or caregiver, for example. That is, the alert display on the display 213 helps a more reliable fall notification to a family member or a caregiver.

  Further, in order to make it easier for surrounding people to notice the fall of an elderly person, an LED (Light Emitting Diode) may be separately provided in the fall determination device 200, and for example, a configuration in which the LED blinks at the time of the fall may be employed. In addition to the above alert display and alert message transmission via the network, the fall determination device 200 may be configured to generate a predetermined alert sound from the speaker 223. The alert sound is effective when, for example, a family member or a caregiver is near an elderly person but is out of sight.

  As described above, according to the present embodiment, when a suspicion of falling is detected, data for a predetermined time thereafter is collected. Then, the candidate of the time of the fall is detected, and the candidate that is highly likely to be related to the motion at the fall is selected from the candidates, and the fall determination is performed using only the displacement amount based on the selected candidate. . In other words, in the stage of obtaining the final displacement amount, candidates that are not likely to be related to the motion at the time of the fall can be well eliminated, so various fall motion patterns ranging from simple fall motion patterns to complex fall motion patterns It is possible to accurately determine whether to fall. In addition, from the viewpoint of performing a fall determination using data for a predetermined time after detection of a suspected fall, it is possible to determine not only simple falls but also various patterns of falls.

  Next, the overturn determination process of the second embodiment will be described. The fall determination process of the second embodiment (and the third embodiment described later) is different from the first embodiment only in the integration range setting process (subroutine in FIG. 9). Accordingly, only the integration range setting process will be described below, and the other processes will be omitted from the above description.

  FIG. 12 shows an integration range setting process subroutine executed in the second embodiment. As shown in FIG. 12, the CPU 203 performs initial value setting (S151). Specifically, the count value i of the built-in counter is set to “1”. Also, the number n of candidate time data acquired in the subroutine of FIG. 6 is set as the maximum value n_max of the count value i. The second embodiment will be described along the example of FIG. Accordingly, the CPU 203 sets “4” as the maximum value n_max.

  Following the processing of S151, the CPU 203 creates an array index of integration start time and integration end time, and sets the table value of the index to “1” (S152). Then, the entry of the integration start time tdi and the integration end time tdi 'corresponding to the candidate time tki is input to the array index corresponding to the current table value (S153). In the second embodiment, the time when the candidate time tki is traced back by time td / 2 is input as the integration start time tdi, and the time when the candidate time tki is advanced by time td / 2 is input as the integration end time tdi ′. That is, an entry indicating the integration range centered on the candidate time is created and input to the array index. The integration time (that is, the time from the start to the end of integration) td applied at this time is a fixed value. The integration time td is, for example, a time obtained as a result of repeated experiments, and is an optimal time for determining whether or not the operation corresponding to the candidate time is an operation related to a fall.

  Following the processing of S153, the CPU 203 determines whether or not the count value i has reached “4”, that is, the maximum value n_max (S154). If the count value i has reached “4” (S154: YES), it is determined that the subroutine of FIG. 12 has been executed for all candidate times, and the process returns to the subroutine of FIG. If the count value i is less than “4” (S154: NO), it is determined in the subroutine of FIG. 12 that there is a candidate time for which the entry of the array index is not input, and the process of S155 is executed. In the process of S155, the CPU 203 increments the count value i and the table value by 1, returns to the process of S153, and repeats the subroutine.

With the subroutine of FIG. 12, four sets of integration start time and integration end time corresponding to the candidate times tk1, tk2, tk3, and tk4 are obtained. Specifically, the following times are obtained as entries in each table. In each table, the former is the integration start time tdi and the latter is the integration end time tdi ′.
[Table 1]... Candidate times tk1-td / 2, tk1 + td / 2
[Table 2] ... Candidate times tk2-td / 2, tk2 + td / 2
[Table 3] ... Candidate times tk3-td / 2, tk3 + td / 2
[Table 4]... Candidate times tk4-td / 2, tk4 + td / 2

  Also in the second embodiment, after setting the integration range, the displacement amount dn corresponding to each candidate time is calculated by applying the integration range. Next, after determining whether or not the displacement amount dn is necessary, the vertical displacement amount Z is calculated using only the vertical displacement amount dn that is highly likely to be included in a series of operations at the time of the fall, and the fall determination is executed. In the second embodiment, the same effect as in the first embodiment is expected.

  Next, the overturn determination process of the third embodiment will be described.

  FIG. 13 shows an integration range setting process subroutine executed in the third embodiment. As shown in FIG. 13, the CPU 203 performs initial value setting (S251). Specifically, the count value i of the built-in counter is set to “1”. Also, the number n of candidate time data acquired in the subroutine of FIG. 6 is set as the maximum value n_max of the count value i. Note that the third embodiment will be described along the example of FIG. Accordingly, the CPU 203 sets “4” as the maximum value n_max.

  Following the processing of S251, the CPU 203 creates an array index of integration start time and integration end time, and sets the table value of the index to “1” (S252). Further, the count value j of the built-in counter is set to “1” (S253).

  Next, the CPU 203 determines whether or not the acceleration absolute value Ai (−j) is equal to or less than the threshold value m (S254). The acceleration absolute value Ai (−j) means the acceleration absolute value j times before the acceleration absolute value Ai corresponding to the candidate time tki. For example, in the case of the absolute acceleration value Ai (−2), it means the absolute acceleration value two times before the absolute acceleration value Ai corresponding to the candidate time tki.

  When the acceleration absolute value Ai (−j) exceeds the threshold value m (S254: NO), the CPU 203 increments the count value j by 1 (S255), and further performs the processing of S254 on the previous acceleration absolute value. Try again. If the acceleration absolute value Ai (-j) is equal to or less than the threshold value m (S254: YES), the time corresponding to the acceleration absolute value Ai (-j) is set as the integration start time tdi, and the entry is the current table value. (S256).

  After resetting the count value j to “1” (S257), the CPU 203 determines whether or not the acceleration absolute value Ai (j) is less than or equal to the threshold value m (S258). The acceleration absolute value Ai (j) means an acceleration absolute value j times after the acceleration absolute value Ai corresponding to the candidate time tki. For example, in the case of the acceleration absolute value Ai (2), it means an acceleration absolute value that is two times after the acceleration absolute value Ai corresponding to the candidate time tki.

  When the acceleration absolute value Ai (j) exceeds the threshold value m (S258: NO), the CPU 203 increments the count value j by 1 (S259), and again executes the process of S258 for the next acceleration absolute value. Execute. When the acceleration absolute value Ai (j) is equal to or less than the threshold value m (S258: YES), the time corresponding to the acceleration absolute value Ai (j) is set as the integration end time tdi ', and the entry is set as the current table value. The corresponding array index is input (S260).

  Following the processing of S260, the CPU 203 determines whether or not the count value i has reached “4”, that is, the maximum value n_max (S261). If the count value i has reached “4” (S261: YES), it is determined that the processes of S253 to S260 have been executed for all candidate times, and the process proceeds to S262. If the count value i is less than “4” (S261: NO), it is determined in S253 to S260 that there is a candidate time for which no entry in the array index is input, and the process of S262 is executed. In the process of S262, the CPU 203 increments the count value i and the table value by 1, returns to the process of S253, and repeats the subroutine.

13 sets of integration start time and integration end time corresponding to the candidate times tk1, tk2, tk3, and tk4 are obtained. Specifically, the following times are obtained as entries in each table. In each table, the former is the integration start time tdi and the latter is the integration end time tdi ′. The numerical value in parentheses means that the integration start or end time is j times before or after the candidate time tki. For reference, FIG. 14 schematically shows the integration start time td and the integration end time tdp ′ corresponding to each table (candidate time).
[Table 1] ... Candidate times tk1 (-1), tk1 (1)
[Table 2] ... Candidate times tk2 (-3), tk2 (2)
[Table 3] ... Candidate times tk3 (-3), tk3 (4)
[Table 4] ... Candidate times tk4 (-5), tk4 (2)

  Here, as shown in FIG. 14, the integration ranges of the candidate times tk3 and tk4 are the same. As described above, when at least a part of the integration range overlaps between the candidate times tki, the integration processing is executed in the subsequent processing. For this reason, the displacement amount Z to be calculated becomes inadvertently large, and the fall determination can be an erroneous determination. Therefore, in the third embodiment, the following processing is executed to merge a plurality of overlapping integration ranges into a single integration range.

  In order to merge overlapping integration ranges, the CPU 203 first sets each parameter value (S263). Specifically, the count value p of the built-in counter is set to “1”, and the number n of candidate time data acquired in the subroutine of FIG. 6 is set as the maximum value n_max (here “4”) of the count value p. To do. Furthermore, the table value of the array index is set to “1” (S264).

  Following the processing of S264, the CPU 203 determines whether or not the integration end time tdp ′ is a time after the integration start time td (p + 1) (S265). When the integration end time tdp ′ is a time after the integration start time td (p + 1) (S265: YES), it is determined that the integration ranges of the candidate time tkp and the candidate time tk (p + 1) overlap, and the process proceeds to S266. . Further, when the integration end time tdp ′ is a time before the integration start time td (p + 1) (S265: NO), the integration ranges of the candidate time tkp and the candidate time tk (p + 1) are not overlapped. Determination is made and the process proceeds to S270.

  In the process of S266, the CPU 203 changes the integration end time tdp ′ corresponding to the current table value to td (p + 1) ′. Next, the count value p and the table value are incremented by 1 (S267), and the table of the corresponding table value is deleted after the increment (S268).

  Following the processing of S268, the CPU 203 determines whether or not the count value p has reached “4”, that is, the maximum value n_max (S269). If the count value p has reached “4” (S269: YES), it is determined that the integration range merging process is necessary and executed, and the process returns to the subroutine of FIG. If the count value p is less than “4” (S269: NO), it is determined that the integration range merging process is not necessary and has not been executed, and S270 is executed. In the process of S270, the CPU 203 increments the count value p and the table value by 1, returns to the process of S265, and repeats the subroutine.

As a result of executing the processing of S263 to S270, the integration ranges corresponding to the candidate times tk3 and tk4 are merged, and the array indexes are as follows.
[Table 1] ... Candidate times tk1 (-1), tk1 (1)
[Table 2] ... Candidate times tk2 (-3), tk2 (2)
[Table 3] ... Candidate times tk3 (-3), tk3 (4)

  Also in the third embodiment, after setting the integration range, the displacement amount dn corresponding to each candidate time is calculated by applying the integration range. Next, after determining whether or not the displacement amount dn is necessary, the displacement amount Z is calculated using only the displacement amount dn that is highly likely to be included in a series of operations at the time of the fall, and the fall determination is executed. In the third embodiment, the same effect as in the first and second embodiments is expected.

  The above is the embodiment of the present invention. The present invention is not limited to these embodiments and can be modified in various ranges. For example, each process may be executed using the absolute value of jerk as an alternative to the acceleration absolute value. In addition, when calculating the amount of displacement from the jerk, the jerk is three-dimensionally integrated with time.

  In another embodiment, a configuration in which the offset value calculation process of S15 in FIG. 5 is not executed may be employed. In this case, resources of the data collection device 100 and the fall determination device 200 can be saved and reduced, and the device can be downsized.

  Moreover, the fall determination device 200 may be configured to execute the main routine of FIG. 5 using a specified offset value according to on / off of an operation switch, for example. In this case, the fall determination device 200 operates in the offset mode only when, for example, the operation switch is turned on. In the offset mode, the overturn determination apparatus 200 performs the process of S16 without executing the process of S15 of FIG. 5, and calculates the vertical displacement amount Z ′ using the specified offset value. According to this embodiment, since it is not necessary to calculate an offset value, an effect such as a processing load on the control system is expected.

  In another embodiment, the data collection device 100 and the fall determination device 200 can be provided as a single device. For example, since all data communication of each component is performed via a bus, an improvement in processing speed is expected.

It is a block diagram which shows the structure of the fall determination system of embodiment of this invention. It is a flowchart of the data collection process performed in the embodiment of the present invention. It is a flowchart of the fall doubt determination process performed in the embodiment of the present invention. It is a graph which shows the relationship between an acceleration absolute value and time, and a displacement amount and time. It is a flowchart of the fall determination process performed in the 1st Embodiment of this invention. It is a flowchart which shows the candidate selection process of S13 of FIG. It is a figure for demonstrating the process of FIG. 6 more concretely. It is a flowchart which shows the displacement amount calculation process of S14 of FIG. It is a flowchart of the integration range setting process of S31 of FIG. It is a flowchart of the offset value calculation process of S15 of FIG. It is a flowchart of the notification determination process of S17 of FIG. It is a flowchart of the integration range setting process performed in the 2nd Embodiment of this invention. It is a flowchart of the integration range setting process performed in the 3rd Embodiment of this invention. It is the figure which expressed typically the integration start and end time corresponding to each candidate time.

Explanation of symbols

1 Fall determination system 100 Data collection device 103, 203 CPU
105, 205 RAM
107 Acceleration sensor 109 Angle degree sensor 111, 211 RTC
113, 207 ROM
115 A / D conversion unit 117, 217 Data transmission / reception unit 119, 219 Power supply 121, 221 Bus 200 Fall determination device 213 Display 215 Communication unit 223 Speaker

Claims (35)

In a fall determination system that determines the fall of a person,
A sensor that is attached to a human trunk and outputs a value corresponding to a vertical displacement of the trunk;
Candidate selection means for selecting an output value satisfying the first condition from among the output values of the sensor for a first time as fall candidate data that can be related to the action at the time of the fall;
Candidate displacement amount calculating means for calculating a candidate displacement amount in the vertical direction of the trunk corresponding to each selected fall candidate data, using an output value of the sensor;
A fall displacement amount calculating means for calculating the fall displacement amount using only the candidate displacement amount satisfying the second condition among the calculated candidate displacement amounts;
A fall determination system comprising: a fall determination unit that determines the fall of the person based on the calculated fall displacement amount. Based on the output of the sensor, further comprising a fall suspect determination means for determining whether or not a suspicious event has occurred,
The said candidate selection means selects fall candidate data for the output value for said first time after the occurrence of the event is detected by said fall suspect determination means. Fall detection system. A correction value calculating means for estimating an error in the output value of the sensor and calculating a correction value;
The fall determination system according to claim 2, further comprising a correction unit that corrects the fall displacement amount based on the calculated correction value.   4. The fall determination system according to claim 3, wherein the correction value calculation means calculates the correction value based on an output value for a second predetermined time immediately before the occurrence of an event that is suspected to fall.   The said correction | amendment means contains the offset value calculation means which calculates the offset value for correct | amending the said amount of fall displacement based on the correction value calculated by the said correction value calculation means, The Claim 3 or Claim characterized by the above-mentioned. Item 5. The fall determination system according to any one of Items 4 to 6.   The fall determination system according to any one of claims 1 to 5, wherein the first condition is to have an output value equal to or greater than a first threshold value.   The said candidate displacement amount calculation means divides | segments the output value for said 1st time for every fall candidate data, and calculates a candidate displacement amount using the output value group of each area. The fall determination system according to claim 6.   The candidate displacement amount calculation means calculates a candidate displacement amount corresponding to each fall candidate data using output value groups for a predetermined time before and after each fall candidate data. The fall determination system according to any one of 6.   The candidate displacement amount calculation means uses each output value group between two output values that are close to each other before and after each fall candidate data and are equal to or lower than a second threshold value that is lower than the first threshold value, to each fall candidate data. The fall determination system according to claim 6, wherein a corresponding candidate displacement amount is calculated.   The candidate displacement amount calculation means determines whether or not the plurality of output value groups include output values in the overlapping period, and determines that the output value group includes the output values in the overlapping period. 10. The fall determination system according to any one of claims 7 to 9, wherein the two are merged into a single output value group.   The fall determination system according to any one of claims 1 to 10, wherein the second condition is a value equal to or greater than a third threshold value.   The fall determination according to any one of claims 1 to 11, wherein the fall displacement amount calculation means calculates the fall displacement amount by adding only candidate displacement amounts satisfying the second condition. system. The sensor includes an acceleration sensor or jerk sensor,
The suspected fall determination means monitors the output of the acceleration sensor or jerk sensor, and determines that an event suspected of falling has occurred when an output value equal to or greater than a predetermined threshold is detected,
The fall determination system according to claim 2, wherein the candidate displacement amount calculation means calculates a candidate displacement amount by time-integrating an output value of the acceleration sensor or jerk sensor. The sensor further includes an angular velocity sensor;
The candidate displacement amount calculation means uses the output value of the angular velocity sensor to convert a vector of output values of the acceleration sensor or jerk sensor so as to correspond to a predetermined static coordinate system, and the converted acceleration sensor The fall determination system according to claim 13, wherein the candidate displacement is calculated by integrating the output value of the jerk sensor with time. Correction selection means for selecting whether or not to correct the fall displacement amount calculated by the fall displacement amount calculation means;
15. The apparatus according to claim 1, further comprising correction means for correcting the fall displacement amount using a predetermined correction value when selected to correct the fall displacement amount. The fall determination system according to any one of the above.   The fall determination system according to any one of claims 1 to 15, further comprising a fall notification means for notifying the fact that the person has fallen by the fall determination means. . The fall notification means
Alert display means that can display alerts,
Alert sound output means for outputting an alert sound,
At least one of alert message sending means for sending an alert message to a predetermined contact;
When the fall determination means determines that the person has fallen, an alert display by the alert display means, an alert sound output by the alert sound output means, an alert to the predetermined contact by the alert message transmission means The fall determination system according to claim 16, wherein at least one of message transmission is executed. In the fall judgment method for judging the fall of a person,
A sensor output step in which a sensor attached to a human trunk outputs a value corresponding to a vertical displacement of the trunk;
A candidate selection step of selecting an output value satisfying the first condition from the output values of the sensor for a first time as fall candidate data that may be related to an action at the time of the fall;
A candidate displacement amount calculating step for calculating a vertical candidate displacement amount of the trunk corresponding to each selected fall candidate data using an output value of the sensor;
A fall displacement amount calculating step for calculating a fall displacement amount using only candidate displacement amounts satisfying the second condition among the calculated candidate displacement amounts;
A fall determination method comprising: a fall determination step of determining a fall of the person based on the calculated fall displacement amount. Further comprising a fall suspect determination step for determining whether or not a suspicious event has occurred based on the output of the sensor;
19. The fall determination method according to claim 18, wherein, in the candidate selection step, fall candidate data is selected based on an output value for the first time after the occurrence of the event is detected in the fall suspect determination step. A correction value calculating step for estimating an error in the output value of the sensor and calculating a correction value;
The fall determination method according to claim 19, further comprising: a correction step of correcting the fall displacement amount based on the calculated correction value.   21. The fall determination method according to claim 20, wherein, in the correction value calculation step, the correction value is calculated based on an output value for a second predetermined time immediately before the occurrence of the suspicious event.   The fall determination method according to claim 20 or 21, wherein, in the correction step, an offset value for correcting the fall displacement amount is calculated based on the correction value calculated in the correction value calculation step. .   The fall determination method according to any one of claims 18 to 22, wherein the first condition is an output value equal to or greater than a first threshold value.   24. Any one of claims 18 to 23, wherein in the candidate displacement amount calculating step, the output value for the first time is divided for each fall candidate data, and the candidate displacement amount is calculated using the output value group of each section. The fall judging method of crab.   24. The candidate displacement amount corresponding to each fall candidate data is calculated using the output value group for a predetermined time before and after each fall candidate data in the candidate displacement amount calculating step. The fall judging method described.   In the candidate displacement amount calculating step, each fall candidate data is obtained by using an output value group between two output values that are close to each other before and after each fall candidate data and are equal to or lower than a second threshold value that is lower than the first threshold value. The fall determination method according to claim 23, wherein a corresponding candidate displacement amount is calculated.   In the candidate displacement amount calculating step, it is determined whether or not the plurality of output value groups include output values of overlapping periods, and when it is determined that the output values of overlapping periods are included, the plurality of output value groups The fall determination method according to any one of claims 24 to 26, wherein the two are merged into a single output value group.   The fall determination method according to any one of claims 18 to 27, wherein the second condition is a value equal to or greater than a third threshold value.   29. The fall determination system according to claim 18, wherein in the fall displacement amount calculation step, the fall displacement amount is calculated by adding only candidate displacement amounts satisfying the second condition. The sensor is an acceleration sensor or a jerk sensor;
In the suspected fall determination step, the output of the acceleration sensor or jerk sensor is monitored and it is determined that an event suspected of falling has occurred when an output value equal to or greater than a predetermined threshold is detected,
The fall determination method according to claim 19, wherein in the candidate displacement amount calculating step, a candidate displacement amount is calculated by time integration of an output value of the acceleration sensor or jerk sensor. The sensor includes an angular velocity sensor,
In the candidate displacement amount calculating step, using the output value of the angular velocity sensor, the vector of the output value of the acceleration sensor or jerk sensor is converted to correspond to a predetermined static coordinate system, and the converted acceleration The fall determination method according to claim 30, wherein the candidate displacement amount is calculated by integrating the output value of the sensor or jerk sensor with time. A correction selection step for selecting whether or not to correct the fall displacement amount calculated in the fall displacement amount calculation step;
The fall according to any one of claims 18 to 31, further comprising a correction step of correcting the fall displacement amount using a predetermined correction value when the fall displacement amount is selected to be corrected. Judgment method.   The fall determination method according to any one of claims 18 to 32, further comprising a fall notification step of notifying the fact that the person has fallen in the fall determination step.   In the fall notification step, when it is determined in the fall determination step that the person has fallen, an alert message is displayed on the display, an alert sound is output by a speaker, and an alert message is sent to a predetermined contact via the network. The fall determination method according to claim 33, wherein at least one of transmitting is performed.   A fall determination program for causing a computer to execute the fall determination method according to any one of claims 18 to 34.

Images