TWI679613B - Method for avoiding false alarm by non-fall detection, an apparatus for human fall detection thereof - Google Patents

Abstract

A false alarm prevention method and a fall detection device through non-fall detection. When a device worn on a person being cared produces a collision signal, the acceleration data is sampled multiple times in a short period of time after the collision signal is generated to stand still. Calculated multiple data to determine whether the device is perpendicular to the ground, to determine whether the event really fell, so it can effectively prevent false alarms.

Description

Method for preventing false alarm by non-fall detection and fall detection device

The specification discloses a method for preventing false alarms in fall detection, particularly a method for preventing false alarms that counts the state of a fall detector in a short time and excludes non-fall events, and related devices.

Human fall detection is one of the most important issues in the care system. Fall detection technology is being developed in care-related fields and can be applied to the care needs of the elderly, frail, and disabled.

The traditional fall detection technology generally uses a wearable device worn on the person being cared for, such as a sensor wristband and a necklace provided to monitor the person's fall.

For example, in the conventional technology, when a caregiver wearing a portable device such as a wristband or a necklace falls, a sensor in the fall is used to detect the fall motion, such as an accelerometer and a gyroscope. ). Generally speaking, an accelerometer is used to measure the change in the acceleration of a portable device. Since the change in acceleration can reflect the change in the posture of the human body, a fall signal is generated when the change in acceleration exceeds a threshold set by the care system.

When the accelerometer recognizes that a larger acceleration is generated by the portable device, it indicates that a collision event is detected. Through the above collision and acceleration changes, the care system The system can accurately know the fall event. In this way, when the care system receives fall, collision and still signals in sequence from the portable device, it will be able to issue an alarm.

The traditional fall detection process is usually based on the change of the acceleration value (a (t)) over time (t), and calculates the total acceleration value (a (t )), Refer to the trend of the total acceleration value (a (t)) as a function of time (t) described in the chart shown in FIG. 1. In the fall detection process, the detection of a fall event can be obtained from the change of the total acceleration value (a (t)) with time (t). For example, the total acceleration value (a (t)) is close to zero. When the total acceleration value (a (t)) increases rapidly, it indicates that it has reached the collision state; when the total acceleration value (a (t)) maintains a stable value, it indicates that it has entered a stationary state. State, these three states become the basis for judging whether there is a fall event.

When a sensor such as an accelerometer in a portable device is sufficiently sensitive, a fall event can be simply sensed through a large change in acceleration value, however, some actions in daily life may also cause the sensor to have a similar fall Therefore, the care system of the portable device that cooperates with the fall detection should need to rule out the possibility of misjudgment.

The specification discloses a method for preventing false alarms through non-fall detection, and a fall detection device, in which a fall detection device event with a sensor is applied to a person being cared for, such as an accelerometer, when there is a collision signal After it is generated, non-falling events are eliminated through the detection of multiple acceleration changes to prevent false alarms and avoid misjudgement of the care system due to error messages, which wastes resources of the care system.

The method for preventing false alarms in this fall detection is mainly applied to a fall detection device worn on a person being cared for. It is determined whether the fall event is a true fall event by sampling the acceleration data from the fall to the standstill, thereby eliminating the possibility of Issue with false alarm.

When the accelerometer in the fall detection device generates a total acceleration value exceeding the threshold At this time, it is judged as a collision event and a collision signal is generated. After that, multiple acceleration data are sampled after receiving the collision signal and before reaching a standstill.

In this process, the total acceleration and the vertical reference axis acceleration sub-vector are calculated based on the three-axis acceleration sub-vectors of each sampled acceleration data. According to the three-axis acceleration sub-vectors, the vertical reference axis acceleration sub-vectors, and Calculate the amount of vertical change in the total acceleration and confirm it against a vertical judgment threshold. This is data that meets non-vertical conditions during the fall. Based on this, you can calculate the amount of vertical change for multiple samples of acceleration data. If the number of non-vertical conditions in China does not reach the proportional threshold, it is confirmed that this is a non-fall event, and the fall detection process is terminated.

In another embodiment, in the step of determining whether the vertical change amount of each pen meets the non-vertical condition, the three-axis acceleration sub-vector of the acceleration data of each sample is sampled, and the total acceleration is calculated, and the acceleration in the fall detection device is calculated. A reference vector pointing to the center of the earth is compared to calculate a vertical change amount, so as to determine whether the speed data meets non-vertical conditions. In the sampled multiple acceleration data, when the number of non-vertical conditions meets a proportional threshold, it is confirmed that this is a non-fall event.

The disclosure also relates to a fall detection device. The main circuit in the device includes a processor, a sensing unit, such as an accelerometer, and a memory unit. The memory unit stores an instruction set. The instruction set is implemented after the processor executes the instruction set. False alarm prevention method with non-fall detection.

In order to further understand the technology, methods and effects adopted by the present invention to achieve the intended purpose, please refer to the following detailed descriptions and drawings of the present invention. It is understood, however, the drawings are provided for reference and description only, and are not intended to limit the present invention.

20,20’‧fall detection device

2 people

40‧‧‧reference vector

50‧‧‧fall detection device

501‧‧‧Micro Control Unit

510‧‧‧gravity direction judgment unit

503‧‧‧sensing unit

505‧‧‧Memory unit

507‧‧‧communication unit

52‧‧‧care system

Step S601 ~ S609‧‧‧ Generate collision signal flow

Steps S801 ~ S811‧‧‧‧ False alarm prevention process through non-fall detection

Figure 1 shows a graph depicting the change in acceleration value a (t) over time in a conventional fall detection process; Figure 2 shows a person wearing a device with a sensor to detect a fall event Schematic diagram of the embodiment; FIG. 3 shows a schematic diagram of an embodiment of a sensor accelerometer in three axial directions in a fall detection device; FIG. 4 shows another embodiment of a triaxial direction of a sensor accelerometer in a fall detection device. Schematic diagram; FIG. 5 shows a main circuit embodiment of a fall detection device with non-fall detection; FIG. 6 shows a schematic diagram of an embodiment that generates a collision signal in a false alarm prevention method by non-fall detection; FIG. 7 shows In the non-fall detection method for preventing false alarms, the acceleration change curve of the fall action is confirmed through multiple sampling values after a collision; FIG. 8 shows a flowchart of an embodiment of the non-fall detection method for preventing false alarms.

The specification discloses a method for preventing false alarms through non-fall detection, and a fall detection device. The main technical feature is that after the collision detection signal is generated by the fall detection device, after a collision and a short period of time before being stationary (such as 1 second, but not limited to this time) Statistics on the state of the fall detection device, and determine whether the device meets the vertical condition by sampling the acceleration data multiple times. When the amount of data that meets the non-vertical condition does not reach a threshold, this can be judged. This event is not a real fall to rule out a non-fall event.

The above-mentioned fall detection device is a wearable device worn by a caretaker. The sensor is an accelerometer to measure the acceleration value of the device over time. The accelerometer may be a three-axis accelerometer. To measure the acceleration vector of three axes, for example, the three acceleration vector of X axis, Y axis and Z axis are respectively a _x , a _y and a _z , and calculate a total acceleration value a accordingly. Generally, we can know The vector sum of the three axial sub-vectors is the total acceleration value. The magnitude of the total acceleration value is equal to the root of the sum of the squares of the magnitudes of each acceleration sub-vector. The formula is as follows:

One of the purposes of the false alarm prevention method and related devices through non-fall detection is to avoid misjudging false alarms for falls. The main mechanism is to determine whether the device meets non-vertical conditions within a period of time from the collision to the time before it is stationary.

FIG. 2 schematically illustrates a fall detection device worn on a care recipient, which has a sensor for detecting a fall event.

The figure shows a fall detection device 20 worn on a person 2 who is a caretaker, such as a patient. The fall detection device 20 is worn to detect his fall. For example, the fall detection device 20 includes a sensor, such as an accelerometer, which measures the acceleration value of the device. The acceleration value represents the rate of change of speed and can be divided into three partial vectors along three axes. . When the fall detection device is in a stable state, such as resting on a desktop, the accelerometer in the fall detection device measures the total acceleration value under the influence of the earth's gravity (g ~ 9.81m / s ² ). When the measuring device is a free-fall body, the measured value of the accelerometer is zero.

FIG. 3 is a schematic diagram illustrating an embodiment of three directions of the sensor in the fall detection device.

The fall detection device 20 includes an accelerometer, which can generate acceleration sub-vector values along three mutually perpendicular axes (X, Y, Z) in real time. The figure shows the three acceleration sub-vector values. The accelerometer can be designed to specify an axial direction as the reference axis pointing to the center of the earth. In one example, the X axis is an axis substantially parallel to the direction of gravity, and is set as a vertical reference axis. The state of this vertical reference axis can be used to determine whether the fall detection device is in a vertical state. In this example, if the X-axis direction of the accelerometer is substantially parallel to the direction of gravity, the acceleration component vector measured on the X-axis is close to the total acceleration, indicating that the fall detection device 20 is in a vertical state.

However, in another embodiment, even if any axis of the accelerometer in the device is not set to point to the center of the earth, a reference vector parallel to the direction of gravity can be specified as a basis for judging whether the device is in a vertical state, as shown in the figure. Fall detection device shown in 4 A schematic diagram of another embodiment of the middle sensor acceleration vector.

FIG. 4 shows an embodiment in which the X-axis of the acceleration in the fall detection device is set on the axis facing the center of the earth different from that described in FIG. 3. The design of the accelerometer in the fall detection device 20 ′ in the embodiment shown in FIG. 4 is different. None of the axes is oriented toward the center of the earth. However, a reference vector 40 can still be set, which is equivalent to the combined vector of the three-axis acceleration sub-vectors of the accelerometer under the normal state (still and normal droop) of the fall detection device 20 '. Therefore, this reference vector 40 can be used as a reference for judging whether the fall detection device 20 'is correctly worn on the care-giver.

The fall detection devices 20 and 20 'described in the above embodiments may preferably be necklace-type detection devices, which can be naturally hung on the body of the care recipient. One of the axes or the set reference vector faces the ground when the device is operating. heart.

FIG. 5 is a diagram illustrating a main circuit of a fall detection device with non-fall detection.

The fall detection device 50 described in the figure is like a wearable device worn on a person being protected. The fall detection device 50 uses an internal sensor (such as an accelerometer) to sense the movement of a person. The detection device is presented, in which the processor receives data from the sensor and executes a fall detection program.

The fall detection device 50 includes a processor, which may be the micro-control unit 501 shown in the figure, for processing the data generated by the sensing unit 503. The sensing unit 503 may be implemented by an accelerometer and coupled to the micro-control unit 501. To measure acceleration data of the fall detection device 50. In one embodiment, an accelerometer, such as a triaxial accelerometer, mounted on the fall detection device 50 is used to measure acceleration sub-vectors in three axes.

In this embodiment, the micro control unit 501 includes a software-implemented gravity direction determination unit 510 for determining whether the fall detection device 50 is in a non-vertical state. For example, when the specific axis (vertical reference axis) or the reference vector of the accelerometer in the fall detection device 50 is not substantially parallel to the direction of gravity, it means that the fall detection device 50 is in a non-vertical state.

The fall detection device 50 includes a memory unit 505, which is coupled to the micro control unit. 501, as the system memory in the device, and is specifically used to store the fall detection program executed by the micro control unit 501, and is particularly used to store an instruction set, which is implemented after being executed by a processor (such as the micro control unit 501). The disclosed false alarm prevention method through non-fall detection, wherein the running program can avoid misjudgment by detecting the state of the fall detection device 50.

In another embodiment, the fall detection device 50 includes a communication unit 507 coupled to the micro control unit 501 to communicate with the care system 52. When the fall detection device 50 detects a fall event through a fall detection method An alarm will be issued and transmitted to the care system 52 through the communication unit 507.

FIG. 6 shows a flow chart (A) of an embodiment in which a collision signal is generated in a false alarm prevention method by non-fall detection.

Step S601 indicates that the sensor of the fall detection device worn on the person being protected continuously generates acceleration data, that is, acceleration sub-vectors of each axis, which have acceleration values and directivity. In step S605, the sensor calculates an acceleration value for determining whether there is a collision based on the acceleration data.

In step S607, the sensor determines whether the calculated total acceleration value is greater than the first threshold. This is a threshold value set by the system to determine whether a collision occurs. That is, in the sensor, the sensor will continuously Sense the axial acceleration sub-vectors and calculate the total acceleration value, and then compare the first threshold to determine whether there is a collision event. If it is not greater than the first threshold, it indicates that it is normal, and the process returns to step S601; If the acceleration value is greater than this first threshold, a collision signal will be generated, as in step S609.

Compared with the traditional method of judging a fall based on changes in the acceleration values of the three stages of fall, collision, and stillness as shown in Figure 1, the method of preventing false alarms through non-fall detection proposed in the disclosure goes further. Whether the fall detection device has a vertical angle change during the fall of the caregiver to eliminate the fall misjudgment.

For example, from the implementation of fall detection, it can be known that when the caretaker actually falls, from the process of hitting the ground, a direction (vertical reference axis or reference vector) set in the fall detection device worn will be changed from Perpendicular to ground instead of non-perpendicular ground.

Conversely, when the care-giver is in normal activities, such as sitting down or bending down, the user occasionally collides (the accelerometer detects a large acceleration value change), but falls within a period of time after the collision, and the fall detection The vertical reference axis or vertical angle of the reference vector and the ground in the device will not change much, and such a situation is not a fall.

It is worth mentioning that when a caregiver lays still on the ground after a fall, the fall detection device worn may still hang next to the body so that the vertical reference axis or reference vector set therein is perpendicular to the ground, so If the wearable device is perpendicular to the ground after the test device is at a standstill, it should not be judged as non-falling.

Therefore, when judging whether to fall through the false alarm prevention method through non-fall detection, the caretaker will be in a period of time (for example, about one Seconds) to determine whether the fall detection device produces a vertical angle change, thereby eliminating misjudgments caused by signals generated by the caregiver during normal activities.

Before describing the flow of FIG. 8, referring to FIG. 7, in a method for preventing false alarms through non-fall detection, the acceleration change curve of the fall action is confirmed through multiple sampling values within a time (such as 1 second) after the collision. .

Figure 7 shows the change of the total acceleration (a (t)) of a fall event over time (t). The first period shows the change in the total acceleration value of the fall process. Then, when the caregiver hits the ground or any The article will produce a larger change in the total acceleration value, that is, the collision displayed in the second period. Then, as in the third period, the fall detection device will not be too large as the caretaker is stationary. The total acceleration value changes.

In the second period of time, that is, from the collision to the standstill, the sub-vector (ax (t)) of the vertical reference axis is significantly different from the total acceleration value, which means that the fall detection device has produced a change in the vertical angle. . Conversely, if it is a signal generated by general activities, the sub-vector (ax (t)) of the vertical reference axis is close to the total acceleration value.

In the third period of time, that is, when the person being cared for still, the fall detection device is suspended beside the body and the vertical reference axis set therein is perpendicular to the ground, so the sub-vector of the vertical reference axis (ax (t )) Close to the total acceleration value.

In the application of the method of preventing false alarms through non-fall detection, because the accelerator will be non-vertical to the ground between the collision and the stationary state, the program running in the processor will use multiple strokes between the collision and the stationary state. Acceleration data, statistically confirm that the vertical reference axis or the reference vector pointing to the center of the earth does reach the non-vertical condition, and avoid using only one or too small amount of data as a basis for excluding misjudgment.

其中a_x，a_y，a_z是三軸加速度計所量得之三軸分向量，a_x是系統設定的垂直參考軸的加速度分向量，a為總加速度。 In the flow (B) of FIG. 8, after receiving the collision signal in step S801, a plurality of pieces of acceleration data are sampled in a period from the collision to the standstill in step S803. Because the caretaker's sensors in the fall detection device will still be disturbed during normal activities, which will cause misjudgment values, multiple samplings are required, and multiple acceleration data are sampled before standing still, as in step S805, respectively Calculate the vertical change based on the sampled acceleration data. The formula for calculating the vertical change a _diff is as follows:

Where a _x , a _y , a _z are the three-axis sub-vectors measured by the three-axis accelerometer, a _x is the acceleration sub-vector of the vertical reference axis set by the system, and a is the total acceleration.

In another embodiment, if none of the acceleration axes is directed to the center of the earth, the system specifies a reference vector (a _xref , a _yref , a _zref ) that points to the center of the earth. This reference vector is when the fall device is stationary and When perpendicular to the ground, the triaxial sub-vector measured by the triaxial accelerometer.

其中a_xref，a_yref，a_zref是系統指定一指向地心的參考向量的三個分向量，a_x，a_y，a_z是三軸加速度計所量得之三軸分向量，a為總加速度。 After operation, the vertical change a _{diff is} as follows:

Wherein _{a xref,} a _yref, a zref specifying a reference vector pointing to the center of the earth three sub-vector _{system, a x, a y, a} z is the amount of three-axis acceleration obtained triaxial vector points, a is the total Acceleration.

It is worth mentioning that the embodiment in which the system specifies a vertical reference axis pointing to the center of the earth is a special case of the reference vector embodiment. In this special case, the value of the reference vector (a _xref , a _yref , a _zref ) is just equal to ( 1,0,0), so the formula of the vertical variation of the reference vector embodiment can be simplified to the formula of the vertical variation with the system-specified vertical reference axis.

Next, it is determined whether the vertical change amount meets a non-vertical condition, as in step S807. In this step, the system can be set as a vertical judgment threshold a _thv for judging whether it is vertical or not. If the vertical variation is larger than the vertical judgment threshold, it is regarded that the device is not perpendicular to the ground: a _diff > a _thv , and Because in order to eliminate the caretaker's normal activity, the sensor will still generate false signals that will cause misjudgment. Therefore, multiple sampling is required, and the number of non-vertical conditions that meet a certain proportion threshold in all sampling values. Think of a fall. In this way, in this fall event, multiple vertical changes are calculated from the sampled multiple sampling values. When the number of times that exceeds the vertical judgment threshold does not reach a certain percentage threshold, it will be considered as not meeting the non-vertical condition at the time of the fall ( No), that is, if step S811 is executed, the current fall detection process is terminated, and the process (A) for judging whether there is a collision signal is initially returned. Conversely, when the number of times that exceeds the vertical judgment threshold reaches the set proportional threshold, it means that enough sampling values are judged to be non-vertical, which also meets the non-vertical condition (Yes), and the system continues to judge other fall conditions (step S809) .

In summary, the above embodiments describe a method for preventing false alarms in fall detection and related devices, which are mainly a method for preventing false alarms that quickly counts the state of the fall detector to exclude non-fall events in a short period of time. The main method is to sample the acceleration data multiple times within a period of time after the collision signal is generated, and to determine whether the event is not a real fall event through judgment of non-vertical conditions, which can effectively prevent false alarms.

The above descriptions are only the preferred and feasible embodiments of the present invention, and any equivalent changes and modifications made in accordance with the scope of patent application of the present invention shall fall within the scope of the present invention.

Claims (9)

A false alarm prevention method through non-fall detection is applied to a fall detection device worn on a person being cared for. The method includes: an accelerometer in the fall detection device generates a collision signal; and the processor receives the Sampling multiple acceleration data within a time after the collision signal but before the device is at rest; judging whether each acceleration data is compared with a gravity direction is not vertical to the ground according to the sampled acceleration data; during the sampling of the multiple acceleration data When the number of non-vertical grounds in the sampled multiple acceleration data does not reach a proportional threshold, the fall detection process is terminated. The method for preventing false alarms through non-fall detection as described in claim 1, wherein the step of generating the collision signal comprises: when the accelerometer generates acceleration data, calculating a corresponding total acceleration value, wherein when the total acceleration value When it is larger than a first threshold, the collision signal is generated. The method for preventing false alarms through non-fall detection according to claim 1, wherein the step of determining whether the comparison between each acceleration data and the direction of gravity is not perpendicular to the ground includes: according to the three axes of each sampled acceleration data The sub-vectors respectively calculate a total acceleration; use the total acceleration and the three-axis sub-vector ratio to calculate a vertical change amount for a vertical reference axis sub-vector of the accelerometer in the fall detection device; and compare the vertical change amount to one The vertical judgment threshold is used to judge whether the speed data is not perpendicular to the ground. The method for preventing false alarms through non-fall detection according to claim 1, wherein the step of determining whether the comparison between each acceleration data and the direction of gravity is not perpendicular to the ground includes: according to the three axes of each sampled acceleration data The sub-vectors respectively calculate a total acceleration, and use the total acceleration and the three-axis sub-vectors to compare and set a reference vector pointing to the center of the earth to obtain a vertical change; and use the vertical change to compare a vertical judgment threshold to judge Whether the pen speed data is not perpendicular to the ground. A fall detection device includes: a processor; a sensing unit is an accelerometer coupled to the processor to generate acceleration data about the fall detection device; and a memory unit is coupled to the processing And a device for storing an instruction set. After the instruction set is executed by the processor, a false alarm prevention method through non-fall detection is implemented to: receive an error generated by an accelerometer in the fall detection device; After the collision signal, multiple acceleration data is sampled within a period of time after receiving the collision signal and before the device is stationary; according to the sampled acceleration data, determine whether the comparison between the acceleration data and a direction of gravity is not vertical to the ground; In the plurality of acceleration data, when the number of non-vertical grounds in the sampled plurality of acceleration data does not reach a proportional threshold, the fall detection process is terminated. The fall detection device as claimed in claim 5, further comprising a communication unit coupled to the processor for transmitting the alarm to a care system. The fall detection device according to claim 5, wherein the accelerometer is a three-axis accelerometer, and a vertical reference axis pointing to the center of the earth is set, or a reference vector pointing to the center of the earth is set. The fall detection device according to any one of claims 5 to 7, wherein, in the method for preventing false alarms through non-fall detection executed by the processor, it is determined whether each piece of acceleration data is compared with the direction of gravity The non-vertical step of the ground includes: calculating a total acceleration according to the three-axis sub-vectors of each sampled acceleration data; and using the total acceleration and the three-axis sub-vector ratio to a vertical reference axis of the accelerometer in the fall detection device. , To obtain a vertical change amount, and use the vertical change amount to compare a vertical judgment threshold to determine whether the speed data is not perpendicular to the ground. The fall detection device according to any one of claims 5 to 7, wherein, in the method for preventing false alarms through non-fall detection performed by the processor, it is determined whether each acceleration data is compared with the direction of gravity The step of non-vertical ground includes: calculating a total acceleration according to the three-axis sub-vectors of each sampled acceleration data; comparing the total acceleration and the three-axis sub-vector to setting a reference vector pointing to the center of the earth to obtain a vertical The amount of change is compared with a vertical judgment threshold with the vertical change, so as to determine whether the speed data is not perpendicular to the ground.