ITTO20101085A1 - DEVICE AND PROCEDURE FOR RECOGNIZING FALSE ALARMS IN A PERSONAL EMERGENCY SIGNALING DEVICE - Google Patents

Description

"Device and procedure for recognizing false alarms in a personal emergency signaling deviceâ €

DESCRIPTION

The present invention relates to a device and a method for recognizing false alarms in a personal emergency signaling device for detecting falls.

Average life expectancy has become increasingly higher thanks to advances in the medical field and an overall better quality of life. The consequent aging of the population, particularly evident in Western society, in turn produces a high demand for home care for elderly people living alone. Falls are one of the factors that most influence the physical and psychological health of the elderly population, as they are the main cause of hospital admissions and deaths related to injuries. Consider, for example, that in the United States one in three adults, with reference to the over-65 category, is subject to an event of this type within a year.

Direct costs attributable to falls were US $ 19 billion in 2000 and direct and indirect costs are expected to exceed US $ 54 billion in 2020. In 2008, also in the United States, over two million people needed to go to an emergency room due to a fall.

The harmful effect of falls and the negative impact on the costs of health services have led to great interest in automatic fall detection systems. The presence of an automatic system capable of detecting falls can improve the care of the elderly.

In fact, non-automatic systems, i.e. those based on explicit activation by the elderly, for example by pressing a button, are generally not satisfactory since the elderly subject to a fall may be the victim of a loss of consciousness. or he may have suffered such trauma that he is unable to seek help.

The most promising approaches are those based on wearable devices that monitor the patient's movements, recognize a fall and report an alarm. Therefore, elderly people living alone are given a Personal Emergency Response System - PERS capable of automatically alerting health facilities in the event of a fall.

Many of the existing devices base their operation on movement data (kinematics and posture information) recorded by accelerometers or gyroscopes positioned in contact with the patient's body.

An initial evaluation of the acquired data can be made by the device itself which is equipped with a processing and calculation unit, while for further processing the collected data are generally sent to a remote station via wireless communication.

However, these devices are subject to the problem of false alarms, since some normal activities of daily life, such as getting up, sitting down, lying down, climbing or descending stairs, are sometimes mistakenly interpreted as falls.

In fact, in order to be effective, such automatic detection devices must offer a high degree of accuracy in detecting real falls. In particular, they must have a good sensitivity (defined as the ratio between the number of falls correctly detected and the falls that have actually occurred) and a good specificity (ability to filter false alarms, defined as the ratio between actions similar to a fall that are eliminated and the total number of actions similar to a fall).

There are some automatic detection devices capable of distinguishing activities of daily life from falls. These devices are based on two different solutions.

In the first case, a fixed kinematic threshold is used, in particular an acceleration or angular velocity value, to verify whether or not there has been a fall. In such systems, however, it is difficult to determine the most correct threshold value: if it is too high, some real falls can be interpreted as normal activities; if it is too low, too many false alarms are generated.

In the second case, a kinematic threshold and patient posture information are combined. In such systems it is necessary to use additional devices to acquire posture data, which are annoying and uncomfortable for patients. Furthermore, it is necessary that the various sensors communicate with each other in order to combine the various information, and this considerably reduces the life of the supply batteries.

The purpose of the present invention is therefore to propose a device and a procedure for the recognition of false alarms, in a personal emergency signaling device, which have a high specificity, which are easy to manufacture and which do not require information. of the patient's posture.

This and other purposes are achieved with a device and a method for recognizing false alarms, the characteristics of which are defined in claims 1 and 4.

Particular manufacturing methods are the subject of the dependent claims, the content of which is to be understood as an integral and integral part of this description.

Further characteristics and advantages of the invention will appear from the following detailed description, carried out purely by way of non-limiting example, with reference to the attached drawings, in which:

- figure 1 is a schematic drawing of a personal emergency signaling device provided with a device for the recognition of false alarms according to the present invention; Figure 2 is a block diagram of a method for recognizing false alarms according to the present invention; And

- figure 3 is a graph of the values of the average of the amplitude variations of the acceleration for a plurality of events.

In figure 1, 1 indicates a personal emergency signaling device designed to be worn by a patient. In particular, it is placed around the waist, for example by means of a belt.

This portable emergency device 1 comprises movement sensors 2, in particular a triaxial accelerometer, adapted to be positioned in contact with the patient's body to detect movement data of said body.

The accelerometer measures linear acceleration values on different x, y, z axes of a Cartesian reference system, but in the device according to the invention only the modulus of the acceleration vector is used (and not, therefore, its three components along the Cartesian axes) as it is not necessary for the personal emergency signaling device 1 to be integral with the patient's body. In this way, any changes in the orientation of the device caused by moving the personal emergency signaling device 1 do not affect the procedure described below.

All the data described below refer to values measured by the motion sensors 2 considering the patient's body immersed in the earth's gravity field. Therefore, the motion sensors 2 will measure, for example, 1 * g (where g is the acceleration of gravity) when the patient's body is stationary, 0 when the body is in free fall, etc.

The motion data are sent by the motion sensors 2 to a processing unit 3 which acquires them and then first calculates the magnitude of the acceleration vector and subsequently performs a series of initial checks, as described below.

The processing unit 3 is connected to a device for the recognition of false alarms 4, which includes at least one decision module 6a, 6b, ..., 6N, designed to receive from the processing unit 3 a information signal 8 representative of the movement data (in particular, the samples of the acceleration module acquired in a predetermined time interval, as specified below), process it as described below and supply it to an output module 10 designed for producing an output signal 12 indicative of whether or not there has been a false alarm.

In the preferred embodiment described below there are two decision modules 6a, 6b but others can also be added, generically indicated with 6N, designed to perform different processing of the movement data in order to increase the accuracy of false alarm recognition. .

The decision modules 6a, 6b, ..., 6N are designed to operate independently and decide whether the trend of the acceleration vector module can be interpreted as characteristic of a daily activity or not.

In the event that the movement data can be interpreted as daily activities, the respective decision module 6a, 6b, ..., 6N emits an intermediate signal 14a, 14b, ..., 14N representative of the fact that it is a fake alarm. For example, the decision module 6a, 6b, ..., 6N associates to a flag a value equal to TRUE in the case of daily activity (false alarm) or equal to FALSE in the case of a probable fall.

The intermediate signals 14a, 14b, ..., 14N are sent to the output module 10, which produces an output signal 12 which indicates a false alarm if at least one of the intermediate signals 14a, 14b, ..., 14N indicates that A false alarm has occurred.

The processing unit 3 is also connected to an emergency recognition module 16 which is set up in a per se known way to receive a warning signal 17 indicating that a probable fall has occurred and send an alarm signal 18 is known per se to a radio interface 20 which is representative of the fact that a fall has occurred. According to the present invention, the emergency recognition module 16 is also arranged to receive the output signal 12 coming from the output module 10 and to send the alarm signal 18 only if the output signal 12 indicates that no it is a false alarm. The emergency recognition module 16, having received the warning signal 17 indicating a probable fall, therefore checks the value of the output signal 12 before sending the alarm signal 18. The radio interface 20 in turn transmits the warning signal. alarm 18 to a remote station so that help can be activated.

Alternatively, the emergency recognition module 16 is integrated in the processing unit 3, which therefore receives the output signal 12 and in turn sends the alarm signal 18.

Alternatively, the device for recognizing false alarms 4 and the processing unit 3 are integrated in a single processing module, which may or may not include the emergency recognition module 16.

A known emergency recognition procedure comprises the steps of detecting motion data representative of a patient's body movement, identifying a probable fall and sending an alarm signal representative of the fact that a fall has occurred.

According to the present invention, the above described emergency recognition process is improved by adding a false alarm recognition step so that the alarm signal is sent only if a false alarm has not occurred.

Figure 2 illustrates a block diagram of the procedure for recognizing false alarms according to the present invention.

As a first operation 100 movement data detected by the movement sensors 2 are acquired and, in step 102, samples of the acceleration module in a first predetermined time interval t1 are calculated on the basis of said movement data.

At step 104 the processing unit 3 checks whether a first predetermined subset of the samples of the acceleration modulus exceeds a first predetermined threshold S1, for example 3 * g. This first subset will be referred to below as the peak that exceeded the first threshold S1.

At the next step 106 it is verified whether the exceeding of the first threshold S1 is followed by a second time interval of predetermined duration t2, for example 1200ms, characterized by a low rate of movement, i.e. in which the acceleration modulus is smaller or equal to a second predetermined threshold S2, for example 2.8 * g. This indicates that the business has ended.

The steps described above are known from automatic detection devices which are based on a fixed kinematic threshold of the type described in the introduction.

In the event that both conditions are satisfied, the processing unit 3 sends to the decision modules 6a and 6b the information signal 8 representative of the samples of the acceleration vector module taken in a third predetermined time interval t3 around the peak which has exceeded the first threshold S1, as specified below. At the same time, the processing unit 3 sends the warning signal 17 to the emergency recognition module 16.

If at least one of the two conditions mentioned above is not satisfied, the procedure is interrupted until the next acquisition of new movement data.

At step 108 the first decision module 6a checks whether the movement data can be traced back to one of the following activities: sit / lie down quickly on a soft / elastic surface or sit quickly on a hard surface.

In the event that the monitored patient sits or lies down quickly on a soft surface such as a bed, sofa or chair, the pattern of the acceleration module is characterized by a peak followed by a series of damped oscillations, with less abrupt changes than those that occur in the event of a fall. This is due to the fact that the kinetic energy is gradually dissipated after the impact.

On the other hand, when the patient sits quickly on a hard surface such as a chair, the variations in the acceleration module can be significant but are then characterized by rapid stabilization.

A fall, on the other hand, is characterized by a violent impact on a hard surface that causes a sudden peak in the graph that represents the trend of the acceleration module over time. In general, such a graph contains a plurality of peaks (although not all greater than the first threshold S1 mentioned above) because different parts of the body touch the ground at different instants of time or because the relatively high kinetic energy causes a sort of â € returnâ € effect on the body or parts of it. The sudden peak is characterized by rapid variations in the modulus of acceleration between one sample and the next.

Verification step 108 is performed by calculating an average of the acceleration amplitude variations (Average Acceleration Magnitude Variation - AAMV) according to the following expression:

acc acc

AAMV = ∠‘i <1> − i

i∈ W number of samples ∈ W

where is it

represents the i-th sample of the acceleration vector, and W is the third time interval t3 mentioned above, that is the time interval that contains the peak that has exceeded the first threshold S1 and a plurality of other samples preceding and following said peak . Advantageously, W has a duration that starts 640ms before the first subset of samples of the acceleration module that have exceeded the first threshold S1 and ends 540ms later.

The value of AAMV is directly proportional not only to how quickly the amplitude of the acceleration changes but also to the number of peaks present in the time interval W.

Figure 3 shows a graph of the AAMV value for a plurality of subsequent events in which a series of real falls (represented with asterisks) and a series of daily activities (represented with squares and triangles) are shown, in which the squares indicate activities of sitting / lying on soft surfaces, while triangles indicate activities of sitting on hard surfaces. Lines A and B indicate the mean values, respectively, of the AAMV values for actual falls and for daily activities.

The value of AAMV, compared with an appropriate reference value, for example 0.27 * g, is therefore used to distinguish daily activities from falls, since real falls have an AAMV value greater than the reference value, while daily activities have an AAMV value that is less than said reference value.

If the AAMV value is lower than the reference value, the first decision module 6a sends an intermediate signal 14a in which the flag is set equal to TRUE (false alarm).

In a preferred variant of the invention, the false alarm recognition process comprises further steps now described.

At step 110 the second decision module 6b checks whether the movement data can be traced back to an activity such as performing a jump or running.

A jump consists of three phases: push, free fall and landing. At step 110 the second decision module 6b recognizes a jump by performing the following steps:

- verify the presence of at least one peak of the acceleration modulus associated with the thrust phase, i.e. the existence of a second subset of samples of the acceleration modulus higher than a third predetermined threshold S3, for example 1.5 * g, preceding the peak that exceeded the first threshold S1;

- calculating a landing instant, such as the instant of time which is at a certain fourth predetermined time interval t4, for example 80ms, before the peak which has exceeded the first threshold S1;

- calculate an instant of end of thrust by looking backwards in the sequence of samples of the acceleration module for a value greater than or equal to a fourth predetermined threshold S4, for example 1 * g, starting the search from a fifth predetermined time interval t5, for example 20ms, before the landing instant;

- calculate a free fall interval as the time interval between the instant of landing and the instant of end of thrust;

- calculate an amplitude of the average free fall acceleration as the average of the acceleration modulus in the free fall interval.

If the free fall interval is greater than a sixth predetermined time interval t6, for example equal to 100ms, and the amplitude of the average free fall acceleration is less than a predetermined fifth threshold S5, for example equal to 0.5 * g, the event is considered a jump and the second decision module 6b sends an intermediate signal 14b in which the flag is set equal to TRUE (false alarm).

As regards the event of a race, being this similar to a sequence of jumps, it is recognized following the same procedure. At this point, at step 112 the output module 10, received the intermediate signals 14a, 14b coming from the decision modules 6a and 6b (or only the intermediate signal 14a coming from the first decision module 6a, in case they are not executed the steps of step 110) verifies whether at least one of said intermediate signals 14a, 14b indicates that a false alarm has occurred and sends a corresponding output signal 12 to the module.

The emergency recognition module 16 therefore sends the alarm signal 18 only if a false alarm has not been signaled.

Alternatively, it is possible to carry out only steps 100, 102, 104, 106 and 110 (and not, therefore, step 108); at step 112 the output module 10 therefore receives only the intermediate signal 14b coming from the second decision module 6b, checks whether this intermediate signal 14b indicates that a false alarm has occurred and sends a corresponding output signal 12 to the module.

In summary, the personal emergency signaling device samples the acceleration with a predetermined frequency, for example 50Hz, and when a possible drop is detected, the output signal 12 is analyzed. If the value of this output signal 12 indicates that a false alarm has not occurred, the event is communicated to a base station or a mobile phone using the radio interface 20 associated with the personal emergency signaling device.

Naturally, without prejudice to the principle of the invention, the embodiments and construction details may be widely varied with respect to what has been described and illustrated purely by way of non-limiting example, without thereby departing from the scope of the invention as defined in the attached claims.

Claims (6)

CLAIMS 1. Device for acknowledging false alarms (4) in a personal emergency signaling apparatus (1) comprising motion sensors (2) adapted to be positioned in contact with the body of a patient to detect movement data of said body, a processing unit (3), connected to the motion sensors (2), able to calculate the modulus of an acceleration vector of said body on the basis of said motion data, an emergency recognition module (16), associated with the processing unit (3), designed to receive a warning signal (17) indicating that a probable fall has occurred, and to send an alarm signal ( 18) representative of the fact that a fall has occurred; the device (4) being characterized in that it comprises: - at least one decision module (6a, 6b, â € ¦6N) designed to receive from the processing unit (3) an information signal (8) representative of the acceleration vector module and to process it in order to supply the module emergency acknowledgment (16) a corresponding intermediate signal (14a, 14b, â € ¦14N) representative of the fact that a false alarm has occurred or not, so that the emergency acknowledgment module (16) sends the alarm signal (18) only if a false alarm has not occurred. Device (4) according to claim 1, comprising: - a plurality of decision modules (6a, 6b, â € ¦6N) designed to each receive from the processing unit (3) an information signal (8) representative of the acceleration vector module and to process it in such a way as to obtain corresponding intermediate signals (14a, 14b, â € ¦14N) representative of the fact that a false alarm has occurred or not; - an output module (10) designed to receive the intermediate signals (14a, 14b, â € ¦14N) and to supply the emergency recognition module (16) with an output signal (12) representative of the fact that a false alarm if at least one of the intermediate signals (14a, 14b, ..., 14N) indicates that a false alarm has occurred. 3. Personal emergency signaling apparatus (1) comprising motion sensors (2) adapted to be positioned in contact with the body of a user to detect movement data of said body, a processing unit (3), connected to the motion sensors (2), able to calculate the modulus of an acceleration vector of said body on the basis of said motion data, an emergency recognition module (16), associated with the processing unit (3), designed to receive a warning signal (17) indicating that a probable fall has occurred, and to send an alarm signal ( 18) representative of the fact that a fall has occurred; the personal emergency signaling device (1) being characterized in that it comprises a device for recognizing false alarms (4) according to claims 1 or 2. 4. Procedure for the recognition of false alarms including the operations of: - acquiring (100) motion data representative of a patient's body motion; - calculating (102) samples of the acceleration module, in a first predetermined time interval (t1), on the basis of said movement data; - verifying (104) whether a first predetermined subset of said samples exceeds a first predetermined threshold (S1); - verify (106) whether the exceeding of the first threshold (S1) is followed by a second predetermined time interval (t2) in which the acceleration modulus is less than or equal to a second predetermined threshold (S2); the procedure being characterized by the fact that it also includes the operations of: - generating an information signal (8) representative of said first subset of acceleration samples; - calculate (108), on the basis of the information signal (8), an average of the amplitude variations of the acceleration according to the following expression: acc MV = <1> ∠’acc AA i i ∠‘i ∈ W numberof elements ∈ W where is it represents the i-th sample of the acceleration module, and W represents a third predetermined time interval (t3) which contains said first subset of acceleration samples plus a plurality of acceleration samples before and after said first subset; - generate a false alarm signal (14a, 14b, â € ¦14N; 12) representative of the fact that a false alarm has occurred if the average of the amplitude variations of the acceleration is lower than a predetermined reference value. 5. Process according to claim 4, further comprising the operations of: - verifying (110), on the basis of the information signal (8), the presence of at least a second predetermined subset of samples of the acceleration modulus higher than a third predetermined threshold (S3) preceding said first subset; - calculating a landing instant such as the instant of time which is at a fourth predetermined time interval (t4) before said first subset; - calculate an instant of end of thrust by looking backwards in the sequence of samples of the acceleration module for a value greater than or equal to a fourth predetermined threshold (S4) starting the search from a fifth predetermined time interval (t5) before the instant of landing; - calculate a free fall interval as the time interval between the instant of landing and the instant of end of thrust; - calculate an amplitude of the average free fall acceleration as the average of the acceleration modulus in the free fall interval; - check if the free fall interval is greater than a predetermined sixth time interval (t6) and if the amplitude of the average free fall acceleration is less than a predetermined fifth threshold (S5) and, if both the above conditions are met, generate a false alarm signal (14a, 14b, â € ¦14N; 12) representative of the fact that a false alarm has occurred. 6. Procedure for the recognition of emergencies including the operations of: - detect movement data representative of the movement of a patient's body; - identify a probable fall; - send an alarm signal (18) representative of the fact that a fall has occurred, the method being characterized in that it comprises the step of recognizing false alarms according to claims 4 or 5 and generating the alarm signal (18) only if no false alarms have been identified.