JP5594724B2 - Fall detection system, fall detection device, fall detection method and program - Google Patents

Description

  The present invention relates to a system, apparatus, method, and program for detecting a fall of a detection target.

  In recent years, the technology to detect the fall of the detection target has quickly detected the fall of a care recipient and the fall / collision of a vehicle, etc., in a passenger support device for a vehicle, in addition to home care, a care assist vehicle, etc. It is attracting attention from the standpoint of issuing requests for nursing care and activating airbags. When the fall target is a human, a plurality of fall modes are assumed corresponding to the cause, state, place, etc. of the fall. In addition, when the detection target is a human, many joints perform a defensive action in preparation for an impact in cooperation with a fall. The fall of the detection target uses various sensors to determine the fall based on the sensor detection value, but as described above, there are various modes of the fall, so how the detection target falls. It is a problem to determine whether or not.

  Conventionally, various systems for detecting a fall of a detection target have been proposed. For example, Japanese Unexamined Patent Application Publication No. 2005-237576 (Patent Document 1) describes a fall determination device, and the disclosed device uses acceleration and angular velocity in three axial directions attached to the trunk of a subject, The post-fall judgment is reduced by calculating the amount of vertical displacement of the trunk and judging the fall. Japanese Unexamined Patent Application Publication No. 2007-151948 (Patent Document 2) describes a fall determination method and apparatus, which detects three-axis accelerations and angular velocities that are attached to the trunk of a subject and are orthogonal to each other. The degree of fall is calculated, and the fall is judged from the value of the degree of jerk.

  Japanese Unexamined Patent Application Publication No. 2009-163538 (Patent Document 3) describes a system for determining a fall with high accuracy even when the fall operation pattern includes various movements. The system described in Patent Document 3 determines whether a sensor that outputs a value corresponding to the amount of vertical displacement of the trunk and an event that is suspected of falling has occurred, and an event that is suspected of falling by the judgment The amount of displacement in the vertical direction of the trunk from the time point when it is determined that the occurrence of the above is calculated, and the fall is determined from the total amount of displacement. Furthermore, Japanese Unexamined Patent Application Publication No. 2007-260389 (Patent Document 4) describes a fall detection device that can detect a fall in a short time and has few false detections. The device described in Patent Document 4 detects an angular velocity and an inclination angle around the axis of the trunk in the front-rear direction, and determines whether to fall.

  Furthermore, fall detection using wearable sensors was reported in M. Kangas et al., “Sensorband fall detector prototype: validation through data collection and analysis revised version”, ISMICT2007, ID1217, December 2007 (Non-patent Document 1). ing. The present inventors also reported fall detection using an acceleration sensor in Y. Enomoto et al., “Data Processing for Fall Detection using an Acceleration Sensor”, ICCCS2009, A4-02, 2009 (Non-patent Document 2). doing.

JP 2005-237576 A JP 2007-151948 A JP 2009-163538 A JP 2007-260389 A

M. Kangas et al., “Sensorband fall detector prototype: validation through data collection and analysis revised version”, ISMICT2007, ID1217, December 2007 Y. Enomoto et al., “Data Processing for Fall Detection using an Acceleration Sensor”, ICCCS2009, A4-02, 2009

  As described above, although various apparatuses and methods for detecting a fall have been proposed, there has been a continuing need to increase the certainty of the fall detection for various fall modes. Furthermore, it has been necessary to perform the fall detection with high accuracy and high speed and to enable efficient fall prediction. Further, there has been a need for a small-sized and low power consumption fall detection device that can be mounted on a nursing care vehicle or the like.

  The present invention has been made in view of the above prior art, and in the present invention, a sensor capable of simultaneously measuring triaxial acceleration and triaxial angular velocity is used, and parameters related to the fall of the detection target are acquired. . In the present invention, the absolute maximum values for the triaxial acceleration and the triaxial angular velocity are tracked in time, and the maximum values in the time sequence of the triaxial acceleration and the triaxial angular velocity are sampled.

  To determine whether to fall, calculate the discriminant value using the maximum acceleration value on the time sequence, and exceed the acceleration threshold value where the discriminant value is set, and each angular velocity threshold value where the maximum angular velocity value is set It is determined in response to exceeding. Further, in another embodiment, the angular velocity is calculated by numerically integrating the angular velocity before and after the time when it is determined that the vehicle has fallen, and the vehicle falls in response to exceeding the angle change threshold set for the angle change. Judging. In the present invention, the fall detection can be performed with high accuracy without using the acceleration in the vertical direction for the determination.

  Furthermore, in another embodiment of the present invention, the fall prediction and the fall judgment are performed using the temporal order and the time interval from the establishment of the fall judgment based on the acceleration data until the maximum value of the angular velocity exceeds the set threshold value. It can be carried out.

  According to the present invention, the fall detection is performed with high accuracy and high speed, the fall prediction can be performed efficiently, and can be mounted on a nursing vehicle or the like, and can be reduced in size and power consumption when mounted on a vehicle. A fall detection system, a fall detection device, a fall detection method, and a program can be provided.

The figure which showed 1st Embodiment of the fall detection system 100 of this embodiment. The figure which showed 2nd and 3rd embodiment of the fall detection system 200 of this embodiment. The figure which showed the functional block 300 of the information processing apparatus 110 of this embodiment. The flowchart of the fall detection method of 1st Embodiment which the information processing apparatus 110 performs. The flowchart of the fall detection method of 2nd Embodiment which the information processing apparatus 110 performs. The flowchart of 3rd Embodiment of the fall detection method. The figure which showed embodiment of the fall detection system 700 in other embodiment of this embodiment. The embodiment of the time sequence of the acceleration data and the angular velocity data acquired according to the present embodiment is divided into a fall event (a) that falls in the traveling direction, a fall phenomenon (b) that falls laterally with respect to the traveling direction, and a traveling direction. The figure shown as a fall event (c) which falls backwards with respect. The table | surface which showed the result of the Example which performed the judgment of the fall event. The table | surface which showed the result of the comparative example which performed the judgment of the fall event.

  Hereinafter, although this invention is demonstrated with embodiment, this invention is not limited to embodiment mentioned later. FIG. 1 shows a first embodiment of a fall detection system 100 according to this embodiment. A fall detection system 100 shown in FIG. 1 includes a sensor 104 attached to or attached to a detection target 102. The sensor 104 can measure the acceleration along the axial direction around the three axes in addition to the angular velocity around the three axes, and the acceleration data and the angular velocity data acquired by the sensor 104 are IEEE802.x or BlueTooth (registered). Can be sent to the fall detection device 108 via a wireless network such as a trademark and a network 106 such as the Internet.

  In the embodiment shown in FIG. 1, the fall detection device 108 is implemented as an information processing device 110 such as a personal computer, a workstation, or a server. An input / output device such as a display device, a keyboard, and a mouse is connected to the information processing device 110, and various types of processing are instructed to the information processing device 110, and processing results of the information processing device can be output externally. Yes.

  The information processing apparatus 110 is a microprocessor of CISC architecture such as PENTIUM (registered trademark), Celeron (registered trademark), Xeon (registered trademark), CORE 2 DUO (registered trademark), PENTIUM (registered trademark) compatible chip, or POWERPC. A RISC architecture microprocessor such as (registered trademark) can be implemented in a single-core or multi-core form. The information processing apparatus 110 is controlled by an operating system such as WINDOWS (registered trademark) 2000, WINDOWS (registered trademark) XP, WINDOWS 7 (registered trademark), UNIX (registered trademark), or LINUX (registered trademark). A program implemented using a programming language such as C, C ++, JAVA (registered trademark), JAVASCRIPT (registered trademark), PERL, and RUBY is executed, and processing for detecting a fall is executed.

  FIG. 2 shows the second and third embodiments of the fall detection system 200 of the present embodiment. 2nd Embodiment and 3rd Embodiment are the forms mounted as a vehicle-mounted sensor, and Fig.2 (a) makes the fall detection apparatus 206 into ASIC, and is integrated in the wheelchair 202 as an integrated module containing the sensor 104. FIG. FIG. 2B shows an embodiment in which the fall detection device 216 is mounted on the electric cart 210 as a module including the sensor 104.

  In the embodiment shown in FIG. 2A, a cared person 204 who needs care is sitting on the wheelchair 202, and the fall detection device 206 of the wheelchair 202 acquires acceleration data and angular velocity data from the sensor 104. Execute the fall detection process. When the fall detection device 206 detects a fall using the acceleration and angular velocity of the wheelchair 202 to be detected, the fall detection device 206 sends a fall notification signal for notifying the fall remotely via the network 106. Send to etc. In addition, the fall detection device 206 can activate support elements such as a seat belt and an airbag.

  In the embodiment shown in FIG. 2B, the electric cart 210 is driven by a driver 212 and moves on a sidewalk using a power system such as a battery / motor. The fall detection device 216 receives the acceleration data and the angular velocity data from the sensor 104 and executes the fall detection process similarly to the embodiment described with reference to FIG. 2A, and is installed remotely via the network 106. The data processing center is notified of the fall. When the electric cart 210 falls, the fall detection device 216 can activate support elements such as the airbag 218 and the seat belt 214 to protect the driver. In addition, it can be mounted as a vehicle-mounted sensor for a vehicle such as a passenger car.

  FIG. 3 shows a functional block 300 of the information processing apparatus 110 of this embodiment. 3 is realized by causing the information processing apparatus to operate the CPU as a functional unit indicated by the functional block 300 by executing the program by operating the CPU. The functional block shown in FIG. 3 is an embodiment when the information processing apparatus 110 is mounted using the information processing apparatus in the embodiment of FIG. The information processing apparatus 110 includes a network interface 302 and a data buffer 304. The network interface 302 receives acceleration data and angular velocity data sent from the network 106 and sends the acceleration data and angular velocity data to the data buffer 304. In the data buffer 304, acceleration data 306 and angular velocity data 308 are stored in association with each sampling time, and are used for the subsequent fall detection processing.

  The information processing apparatus 110 further includes an acceleration maximum value extraction unit 310. The acceleration maximum value extraction unit 310 reads the acceleration data 306 stored in the data buffer 304, and extracts the maximum value in absolute value from the axial translational acceleration along the directions of the roll axis and the pitch axis. The maximum value can be extracted using various algorithms, but the maximum value (absolute value) of acceleration in the roll axis and pitch axis directions corresponding to a specific sampling time is registered in the buffer memory. If the maximum value (absolute value) of the three-axis acceleration values at the sampling time is larger than the previously registered value, a process of updating the value registered in the buffer memory can be used. FIG. 3 shows the definition of the coordinate axis system in the present embodiment.

  Further, in the embodiment to be described, the acceleration maximum value extraction unit 310 is described as an axis that is orthogonal to the traveling direction and the traveling direction each time the value in the buffer memory is updated in either the roll direction or the pitch direction. On the other hand, the buffered acceleration of the roll axis and the pitch axis is sent to the fall determination unit 318. In addition, the buffer memory that stores the current maximum value is reset periodically after detecting a fall event or a sudden spike signal due to the influence of noise or the like, and can always properly detect a fall. To be maintained.

  The fall determination unit 318 substitutes the absolute value in the biaxial direction excluding the received acceleration in the direction along the vertical direction (yaw axis) into the discriminant in which the coefficient is determined by multivariate regression analysis. The general form of the discriminant is shown in the following formula (1).

In the above formula, a, b, and c are regression coefficients, x ₁ is an absolute value of acceleration along the traveling direction (yaw axis), and x ₂ is an absolute value of acceleration along the pitch axis. Yes, y is a discriminating value indicating an expected fall value, and takes a value from 0 (no fall) to 1 (fall). The regression can be performed using a well-known technique that gives a minimum residual, but any other method can be used. The yaw axis can be uniquely determined by mounting the sensor 104 so as to coincide with the normal traveling direction of the detection target. The fall determination unit 318 determines that the fall is in the acceleration scale when the discriminant value calculated by the discriminant has a value of 0.5 or more.

  Furthermore, the information processing apparatus 110 includes an angular velocity maximum value extraction unit 314. The angular velocity maximum value extraction unit 314 includes the values of the angular velocity set around three axes at a specific sampling time recorded in the data buffer 304. The maximum absolute value is extracted from the angular velocity data 308 and stored in the buffer memory. Thereafter, the maximum angular velocity value extraction unit 314 compares the maximum absolute value in the angular velocity data set at the subsequent sampling time with the value registered in the buffer memory, and the maximum absolute value of the angular velocity set at the subsequent time is determined. If so, the buffer memory value is updated with the new maximum value. When the value is updated, the updated value is sent to the fall determining unit 318, and the fall determining unit 318 determines whether or not the angular velocity exceeds the angular velocity threshold determined by the multivariate analysis.

  In addition, the fall determination unit 318 includes an angle change integration unit 312. When the fall change integration unit 318 determines that the fall has occurred using the acceleration as a scale, the most of data processed at that time is the most. For an axis having a large angular velocity, the angular velocity is numerically integrated in the range of front and rear Δt around the sampling time T, and the change in angle between Δt is calculated. The fall determination unit 318 compares the calculated angle change with a threshold value obtained by multivariate regression analysis, and if the angle change between Δt is equal to or greater than the threshold value, the fall determination unit 318 determines the fall using the angle change as a scale. .

  As described above, in this embodiment, it is possible to adopt the rotational motion of the trunk perpendicular to the vertical direction as a measure of the fall determination instead of using the vertical acceleration data without using the vertical acceleration change. And For this reason, when the detection target is a human, the possibility of misjudging a vertical movement in a normal action such as squatting or sitting in a chair as a fall is minimized, and measurement accuracy is significantly improved. Can be made.

  The information processing device 110 registers data used for determining the fall, such as angular velocity data, acceleration data and determination values, angular velocity, and angle change, in a persistent storage device such as the HDD device 316, and performs subsequent actions. It can be provided for log analysis.

  FIG. 4 is a flowchart of the fall detection method according to the first embodiment executed by the information processing apparatus 110. The process of FIG. 4 starts from step S400, and in step S401, acceleration data and angular velocity data are acquired. In step S402, the absolute maximum value in the triaxial acceleration data in the time sequence is sequentially extracted, and the maximum value is updated. Every time it is done, the absolute maximum value of acceleration is sent to the fall determination unit 318.

  In step S403, when the value of the buffer memory is updated in the time sequence, acceleration data along the roll axis and the pitch axis excluding the vertical direction is acquired as a discriminant, and the discriminant expressed by the above formula (1) is obtained. Substituting into, calculate the discriminant value. In step S404, the discriminant value is compared with a threshold value (0.5). If the discriminant value is equal to or greater than the threshold value (yes), it is determined that the vehicle falls and the process is passed to step S405. On the other hand, if it is determined in step S404 that the discriminant value is equal to or less than the threshold value, the process returns to step S401 as non-falling (no), and subsequent acceleration data on the time sequence is acquired and the process up to step S404 is performed. Repeat the process.

  In step S405, the fall determination on the angular velocity scale is continuously performed, the triaxial angular velocity data is obtained from the data buffer 304, and compared with the past maximum angular velocity stored in the buffer memory, at the currently determined sampling time. If the absolute value of the angular velocity is larger than the buffered value, the value of the buffer memory is updated. In step S406, it is determined whether or not the updated maximum value of the angular velocity exceeds a threshold value. If the maximum value is exceeded (yes), the process is passed to step S407. On the other hand, if it is determined that the maximum value of the updated angular velocity is not exceeded (no), the process returns to step S401, and the subsequent data processing is repeated.

  In step S407, since the judgments in steps S404 and S406 both indicate a fall, a signal notifying the detection target fall event is output to the outside. In step S408, it is determined whether or not the judgment should be terminated. To do. In step S408, if a stop command for the overturn detection process is issued (yes), the process is stopped in step S409, and otherwise (no), the process returns to step S401 until the measurement and the process are completed. The processing sequence from S401 to step S408 is repeated.

  FIG. 5 is a flowchart of the fall detection method according to the second embodiment executed by the information processing apparatus 110. The second embodiment is a fall detection process using three criteria of an acceleration scale, an angular velocity scale, and an angle change scale. The processing shown in FIG. 5 uses the same sequence as that of the first embodiment shown in FIG. 4 from step S501 to step S506, and thus detailed description up to step S506 is omitted.

  If it is determined that the maximum value of the angular velocity extracted in step S506 exceeds the threshold value (yes), control is passed to step S507, and the sampling time T at which the maximum value of acceleration is updated is acquired. In step S508, the calculated angular change is obtained by multivariate regression analysis by numerically integrating the angle changed around the coordinate axis with the maximum angular velocity over Δt before and after the calculation, and calculating the maximum angular change in the range of Δt. It is determined whether or not the threshold value is exceeded, and if the angle change exceeds the threshold value (yes), control is passed to step S509. On the other hand, if it is determined in step S508 that the angle change does not exceed the threshold value (no), the process returns to step S501 to repeat the process.

  In step S509, a signal notifying that a fall event has occurred is output to the outside, and in step S510, it is determined whether or not the determination should be terminated. In step S510, if a stop command for the overturn detection process has been issued (yes), the process is stopped in step S511, and otherwise (no), the process returns to step S501 until the measurement and the process are completed. The processing sequence from S501 to step S510 is repeated.

  In the second embodiment shown in FIG. 5, by adding the maximum angle change over Δt to the fall determination, it becomes possible to add the angle change of the trunk of the detection target to the determination criterion, and the detection accuracy is further improved. Can be improved. Note that neither the first embodiment shown in FIG. 4 nor the second embodiment shown in FIG. 5 uses vertical acceleration, so that normal operations such as squatting and sitting are erroneously detected. It is possible to minimize the detection and to make the fall detection more efficient. Furthermore, in the second embodiment, the change in the angle of the trunk can be added as a determination criterion, so that even if the detection target simply performs a quick action or performs a spin motion, it may be erroneously determined to fall. As a result, it is possible to perform the fall detection with higher accuracy.

  FIG. 6 shows a third embodiment of the fall detection method. The third embodiment is the same as the first embodiment in that both the fall determination on the acceleration scale and the fall determination on the angular velocity are used, but the determination on the angular velocity scale is performed in advance, and then the acceleration is performed. Judgment is made on a scale, and falls are judged using a time interval until judgment on an angular velocity and acceleration scale is made. According to the third embodiment, in the measurement experiment of the fall event by the present inventors, it is found that the profile indicating the fall event of the time sequence of the maximum value of the angular velocity is not delayed from the profile of the time sequence of the maximum value of the acceleration. It was made by.

  On the other hand, for the same fall event, acceleration and angular velocity peaks occur close in time. Therefore, by using the time difference between the time when the fall event occurred on the acceleration scale and the time when the fall event occurred on the angular velocity scale, the correlation between angular velocity and acceleration is given, and fall detection using both angular velocity and acceleration is used. Can improve the accuracy.

  There are various reasons for this, but at the stage of falling, the coordinate velocity orthogonal to the direction of falling (the direction in which the trunk moves) increases the angular velocity around the coordinate axis. Acceleration changes greatly. When the falling phenomenon proceeds in various manners, the acceleration directed in the falling direction becomes larger than the acceleration of the other axes, and as a result, in this embodiment, the time sequence of the angular velocity is more temporal than the acceleration directed in the falling direction. It can be considered that this is because a peak is formed prior to, and the acceleration related to the falling direction gives a maximum value after a slight delay.

  The process of FIG. 6 is the same as the first and second embodiments in that the fall determination is performed using the biaxial acceleration in steps S600 to S604, and thus detailed description thereof is omitted. If it is determined in step S604 that a fall event has occurred (yes), it is determined in step S605 whether the maximum value of the three-axis acceleration in the time sequence exceeds a threshold value. When the maximum value of acceleration exceeds the threshold value (yes), as a fall event occurs, a time difference ΔT between the time when the discriminant value exceeds the threshold value and the time when the angular velocity exceeds the threshold value is calculated in step S607.

  In step S608, it is determined whether the angular velocity exceeds the angular velocity threshold at the same time or at an early timing, and ΔT is equal to or smaller than the time threshold. If ΔT is equal to or smaller than the time threshold value (yes), a signal notifying the occurrence of a fall event is output in step S609, and it is determined in step S610 whether or not the determination should be terminated. In step S610, if a stop command for the overturn detection process has been issued (yes), the process is stopped in step S611, and otherwise (no), the process returns to step S601 until the measurement and the process are completed. The processing sequence from S601 to step S611 is repeated. It should be noted that in yet another embodiment targeting a wider range of falling events including human walking, it is possible to make a fall determination simply by using the criterion that the time difference ΔT is equal to or less than the time threshold.

  The third embodiment described with reference to FIG. 6 uses the timing in the time sequence of the angular velocity and acceleration as a determination measure instead of the angle change in the second embodiment, can improve the accuracy, and can determine the real time of the determination. It is possible to improve the performance, and further, it is possible to prevent misjudgment due to elimination of influences such as noise.

  FIG. 7 shows an embodiment of a fall detection system 700 in another embodiment. The fall detection system 700 shown in FIG. 7 includes a sensor 710, an A / D converter 720, and a fall detection device 730. Note that the fall detection system 700 may be implemented as a sensor board configured as an integral unit. The sensor 710 samples the acceleration in the direction along the coordinate axis and the angular velocity around the coordinate axis at a sampling frequency of 100 Hz, and sends it to the A / D converter 720. The A / D converter 720 digitally converts the received signal and sends the acceleration data and angular velocity data to the fall detection device 730. The fall detection device 730 is configured as a semiconductor device such as an ASIC, and the semiconductor device is realized as a functional unit illustrated in FIG. 7 by executing a program language such as an assembler language. The fall detection device receives the digitally converted data in association with the sampling time, and stores it in the data buffer 731.

  The data buffer 731 stores the acceleration data and the angular velocity data together with the sampling time, and sends the angular velocity data and the acceleration data to the acceleration maximum value extraction unit 734 and the angular velocity maximum value extraction unit 736. The acceleration maximum value extraction unit 734 and the angular velocity maximum value extraction unit 736 perform the same processing except that the processing described with reference to FIG. Further, the angle change integration unit 735 integrates the angular velocity over the time Δt before and after the sampling time T determined to be the fall event occurrence on the acceleration scale, calculates the angle change before and after the fall event occurs, and falls Send to 737. The fall determination unit 737 executes the same process as described with reference to FIG. 3 except that the process described with reference to FIG. It is sent to the device as a bus line or TCP packet.

  The fall detection system 700 of the embodiment shown in FIG. 7 can be preferably used in the embodiment shown in FIG. 2, and the fall detection device 730 receives a signal notifying the occurrence of the fall event, and the airbag or seat belt. It can utilize suitably in order to start support members, such as.

  The present invention has been described with the embodiments, but the present invention will be described more specifically with the following examples.

  A. Regression analysis-Volunteer subjects (gender: male, age 21-27 years old, height: 163 cm-183 cm, weight: 58-80 kg) recruited 11 wearable sensors that can measure triaxial acceleration and triaxial angular velocity (Hitachi Metal Co., Ltd., wireless-T, sampling frequency 100 Hz) was attached. There are four types of tipping movements (front, side, and rearward tilting movements, and forward tilting movements), and non-falling movements are assumed in ADL (Activities of Daily Living) indoors. 6 types of walking (flat-walking, stair-climbing, sitting on a chair, lying on a futon, carrying a baggage, holding off a fall) were measured at least 5 times per subject and used as data for regression analysis.

  The discriminant and the threshold values obtained by the regression analysis are shown below.

  B. Fall detection-Using data obtained from wearable sensors, data sent from wearable sensors via Bluetooth protocol was obtained, data analysis was performed using a fall detection program installed on a personal computer, and obtained by regression analysis A fall decision was made based on the comparison with the threshold. For the regression analysis method, refer to Non-Patent Document 2 for details. As an example of determining a fall event, a method using an acceleration scale, an angular velocity scale, and an angle change scale is used. The angle change is based on a sampling time T determined as a fall event on the acceleration scale, and before and after the sampling time T 1 The angular velocity of the second range was integrated, the change in angle was calculated, regression analysis was performed, and the correct answer rate of falling judgment was taken as an example.

  C. Comparative Example-As a comparative example, with the correct answer rate of the fall determination for the same data set, except that the fall decision was made by the method described in Non-Patent Document 2 where the fall decision was made using the acceleration in the triaxial direction. did. The discriminant and the threshold values obtained by the regression analysis are shown below.

  FIG. 8 shows an embodiment of a time sequence of acceleration data and angular velocity data acquired according to the present embodiment, a fall event (a) that falls in the traveling direction, a fall phenomenon (b) that falls laterally with respect to the traveling direction, And it shows as a fall event (c) which falls backward with respect to the advancing direction. In FIG. 8, the units of the horizontal axis and the vertical axis are representative of FIG. As shown in FIG. 8, in any case, the vicinity of the time region where the acceleration and the angular velocity form a peak corresponds to the falling event as compared with the case of the non-falling motion. As shown in FIG. 8, in any falling event, the angular velocity tends to increase at a relatively early stage and then the acceleration starts to increase, and the time difference ΔT is approximately the same or less than about 0.5 seconds. It is shown that In FIG. 8, data other than the fall event corresponds to data during walking.

  The results of determining the fall event using these data are shown in the table of FIG. 9 and the comparative example as a table of FIG. In FIG. 10, the normal operation means a correct answer rate for a non-falling phenomenon such as walking. The results shown in FIG. 9 and FIG. 10 are for the purpose of regressing the individuality of every 11 subjects, and 11 combinations of subjects were created, and the regression coefficient was calculated as the 11 average values. The combination is “Number of subjects”, and the correct answer rate and average value for the fall events in three directions are shown.

  As shown in FIG. 9, in this embodiment, it was shown that the correct answer rate was 100% in any case. Moreover, about the result of the comparative example shown in FIG. 10, the correct answer rate of about 97% was given, and it was shown that there is dependence on the direction of the fall event. As a result, in this embodiment, it was shown that the correct answer rate did not change in any direction compared to the comparative example, and the fall event could be detected with high accuracy.

  As described above, the present invention has been described with embodiments and examples in order to facilitate understanding of the present invention. However, the present invention is not limited to the described embodiments and examples, and the operational effects of the present invention. Other embodiments, modifications, and changes that can be conceived by those skilled in the art are included in the scope of the present invention as long as the above is achieved.

  The above functions of the present invention can be implemented in an object-oriented programming language such as C, C ++, Java (registered trademark), JavaScript (registered trademark), Perl, Ruby, or an assembler language, and is a device-readable recording medium. It can be stored and distributed or transmitted and transmitted.

DESCRIPTION OF SYMBOLS 100 ... Fall detection system 102 ... Detection target 104 ... Sensor 106 ... Network 108 ... Fall detection device 110 ... Information processing device 200 ... Fall detection system 202 ... Wheelchair 204 ... Care receiver 206 ... Fall detection device 210 ... Electric cart 212 ... Driving Person 214 ... Seat belt 216 ... Fall detection device 218 ... Air bag 300 ... Functional block 302 ... Network interface 304 ... Data buffer 306 ... Acceleration data 308 ... Angular velocity data 310 ... Acceleration maximum value extraction unit 312 ... Angular change integration unit 314 ... Angular velocity Maximum value extraction unit 316 HDD device 318 Fall determination unit 700 Fall detection system 710 Sensor 720 A / D converter 730 Fall detection device 731 Data buffer 732 Acceleration data 733 Angular velocity data 734 ... Maximum acceleration value extraction unit 735 ... Angular change integration unit 736 ... Angular velocity maximum value extraction unit 737 ... Fall determination unit

Claims (12)

A fall detection system for detecting a fall of a detection target, the fall detection system,
A sensor that is mounted on the detection target and detects a translational acceleration in three axial directions and an angular velocity around the three axes with respect to the posture of the detection target;
A fall detection device that receives translational acceleration data and angular velocity data from the sensor,
The fall detection device
An acceleration maximum value extracting means for receiving the translation acceleration data and the angular velocity data, extracting translation acceleration data having the maximum absolute value of the translation acceleration data for the two axes at the sampling time, and updating the maximum value of the translation acceleration;
Angular velocity maximum value extraction means for extracting angular velocity data having the maximum absolute value of angular velocity at the sampling time from the angular velocity data and updating the maximum value of angular velocity;
A fall determination means for calculating an expected fall value using the extracted translational acceleration data for the two axes , and determining that the detection target has fallen when the expected value exceeds a set threshold value. Fall detection system including and. Further, the fall determination means calculates an angular change around the axis having the maximum motion speed by time-integrating the angular speed data over and around the time when the translation acceleration data exceeds the maximum value. determining a fall of the detection target by angle change exceeds the angle change threshold, fall detection system according to claim 1. Further, the overturn determination means is configured such that the time difference between the time when the translational acceleration data and the angular velocity data exceed the maximum values and the translational acceleration data exceeds the maximum value and the time when the angular velocity data exceeds the maximum value. The fall detection system according to claim 1, wherein the fall of the detection target is determined when at least the time difference is equal to or less than a set time threshold value.   The fall detection system according to any one of claims 1 to 3, wherein the sensor is a wearable sensor or an in-vehicle sensor. A fall detection device for detecting a fall of a detection target,
An acceleration maximum value extracting means for receiving the translation acceleration data and the angular velocity data, extracting translation acceleration data having the maximum absolute value of the translation acceleration data for the two axes at the sampling time, and updating the maximum value of the translation acceleration;
Angular velocity maximum value extraction means for extracting angular velocity data having the maximum absolute value of angular velocity at the sampling time from the angular velocity data and updating the maximum value of angular velocity;
A fall determination means for calculating an expected fall value using the extracted translational acceleration data for the two axes , and determining that the detection target has fallen when the expected value exceeds a set threshold value. Fall detection device including and. Further, the fall determination means calculates an angular change around the axis having the maximum motion speed by time-integrating the angular speed data over and around the time when the translation acceleration data exceeds the maximum value. The fall detection device according to claim 5, wherein a fall of the detection target is determined when an angle change exceeds an angle change threshold set for the angle change. Further, the overturn judging means is configured such that the time difference between the time when the translational acceleration data and the angular velocity data exceed the maximum values and the translational acceleration data exceeds the maximum value and the time when the angular velocity data exceeds the maximum value. The fall detection device according to claim 5, wherein the fall of the detection target is determined when at least the time difference is equal to or less than a set time threshold value.   The sensor is a wearable sensor or an in-vehicle sensor, and the fall detection device is an information processing device that receives data from the sensor by wireless communication or a module that is attached to the detection target together with the sensor. The fall detection device according to any one of? 7. A fall detection method for detecting a fall of a detection target, the fall detection method,
Detecting a translational acceleration in three axial directions and an angular velocity about three axes with respect to the posture of the detection target, attached to the detection target by a sensor;
A fall detection device receiving translational acceleration data and angular velocity data from the sensor;
Receiving the translation acceleration data and the angular velocity data, extracting the translation acceleration data having the maximum absolute value of the translation acceleration data for the two axes at the sampling time, and updating the maximum value of the translation acceleration;
Extracting the angular velocity data having the maximum absolute value of the angular velocity at the sampling time from the previous angular velocity data and updating the maximum value of the angular velocity ;
Calculating an expected fall value using the extracted translational acceleration data for the two axes , and determining that the detection target has fallen when the expected value exceeds a set threshold value ; Including fall detection methods. further,
Calculating the change in angle around the axis with the maximum motion speed by time-integrating the angular speed data over and before and after the time when the translational acceleration data exceeds the maximum value ;
Determining the fall of the detection target when the angle change exceeds an angle change threshold set for the angle change;
The fall detection method of Claim 9 containing these. further,
Calculating the time difference between the time when the translational acceleration data and the angular velocity data exceed each of the maximum values and the translational acceleration data exceeds the maximum value and the time when the angular velocity data exceeds the maximum value ;
The fall detection method according to claim 9, further comprising: determining whether the detection target falls when at least the time difference is equal to or less than a set time threshold value. The computer-executable program for an apparatus to perform the fall detection method of any one of Claims 9-11.