WO2019033235A1

WO2019033235A1 - Fall detection system

Info

Publication number: WO2019033235A1
Application number: PCT/CN2017/097370
Authority: WO
Inventors: Michael Austria ALBANO; Teodoro Angelo San Juan UMALI; Joseph Ian Salvador BALUCAN
Original assignee: Simple Wearables Limited
Priority date: 2017-08-14
Filing date: 2017-08-14
Publication date: 2019-02-21

Abstract

A fall detection system (1) comprising one or more movement sensors (2) for detecting a movement of a user. A processor (3) is provided for receiving movement data from the movement sensors (2). The processor (3) analyzes the movement data to identify a free-fall (4) by the user followed by an impact (5) by the user. The processor (3) identifies a detected fall and sends an alert signal when the free-fall (4) reaches a free-fall threshold (6) and the impact (5) reaches an impact threshold (7).

Description

Fall Detection System

The invention relates to fall detection systems, particularly fall detection systems with portable sensors worn by a user. The present invention is described herein primarily in relation to, but is not limited to, fall detection systems that can be remotely monitored.

Fall detection is important and sometimes crucial for many users, situations, and circumstances. For example, fall detection is important for the elderly, infirm, injured, disabled, and otherwise physically impaired and vulnerable to falls, especially those who live alone or are left alone for extended periods. Fall detection is also applicable in for those who work in hazardous environments such as chemical, industrial, and manufacturing plants, especially those who work alone for extended periods.

Prior fall detection devices use sensors such as accelerometers, gyroscopes, altimeters, and proximity sensors to detect whether a user has fallen. Various algorithms have been used to analyze data gathered by such sensors to determine whether a fall has occurred. However, one major problem with prior devices is the elimination of false positives, where the sensors detect movement that is determined as a fall by the device when in fact there was no fall.

For example, a fall detection device based on accelerometer data can detect when a large impact has occurred and characterize this as a fall. However, the same large impact can result from a user jumping up and down or jumping from a height. A device based on an altimeter can detect a predetermined change in vertical height and characterize this as a fall. However, the same change in vertical height can result from a user lying down of her/his own volition.

If false positives occur too often, users, their carers, and emergency responders will be dissuaded from using the devices for fear of wasting valuable time, resources, and costs in attending to false alarms. Over time, this may adversely affect the sense of urgency in attending to fall alerts if a significant number of alerts turn out to be false alarms.

Fall prediction is also important as a preventative measure before a fall occurs. Fall prediction devices typically monitor one or more particular movement parameters of a user to detect any changes that may indicate an increased susceptibility to falling. Clinical intervention can then be ordered to assess a user and to implement or modify any remedial actions that may be deemed necessary.

However, due to the complex nature of human movement and the many and varied contexts of human movement, it has been difficult to select which movement parameters and what algorithms to use to reliably predict falls.

Mobility tests have also been employed to assess fall risk. However, this requires an investment of time and cost for a test subject, including the engagement of clinicians and other health care professionals. This can be inconvenient, especially if frequent or regular tests over an extended time are required.

It is an object of the present invention to overcome or ameliorate at least one of the disadvantages of the prior art, or to provide a useful alternative.

An embodiment of the present invention provides a fall detection system comprising:

one or more movement sensors for detecting a movement of a user; and

a processor for receiving movement data from the movement sensors, the processor analyzing the movement data to identify a free-fall by the user followed by an impact by the user, and the processor identifying a detected fall and sending an alert signal when the free-fall reaches a free-fall threshold and the impact reaches an impact threshold.

In one embodiment, the processor analyzes the movement data to identify a timeout period beginning after the free-fall, the processor only sending the alert signal when the impact occurs within the timeout period.

In one embodiment, the processor analyzes the movement data to identify a post-fall period beginning after the impact, the processor only sending the alert signal when no further movement reaches a post-fall threshold within the post-fall period.

In one embodiment, one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on previously recorded movement data.

In one embodiment, one or more movement parameters are identified from previously recorded movement data. In one embodiment, the movement parameters comprise one or more of the following: sit-to-stand time; walk speed; turn-around time; and stand-to-sit time.

In one embodiment, a simulated timed-up-and-go (TUG) test is compiled from the movement parameters to calculate a TUG score.

In one embodiment, the movement parameters comprise one or more of the following: sit-to-stand jitter; sit-to-stand delay; walking delay; walking jitter; turn-around delay; turn-around jitter; stand-to-sit jitter; stand-to-sit delay; cadence; gait speed; step length; and step impact.

In one embodiment, a fall risk score is calculated based on one or more of the movement parameters.

In one embodiment, one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on one or more movement parameters identified from previously recorded movement data.

In one embodiment, false positive data and/or true positive data are identified from previously recorded movement data. In one embodiment, one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on the false positive data and/or true positive data. In one embodiment, the detected fall is compared with the false positive data and/or true positive data before sending the alert, and the alert is only sent when the detected fall is determined to be a true positive based on the false positive data and/or true positive data.

In one embodiment, one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on one or more personal parameters of the user. In one embodiment, the personal parameters comprise one or more of the following: age; gender; fall history; medical history; injury history; and blood pressure.

In one embodiment, the fall detection system comprises an analytics module for assessing the risk of a fall and/or for predicting a fall based on previously recorded movement data.

In one embodiment, the movement sensors comprise one or more of the following: an accelerometer; a 3-axis accelerometer; a 6-axis accelerometer; a 9-axis accelerometer; a multi-axis accelerometer; a gyroscope; a 3-axis gyroscope; and an altimeter.

In one embodiment, the fall detection system comprises one or more supplemental sensors. In one embodiment, the supplemental sensors comprise one or more of the following: a temperature sensor; a heart rate sensor; and a blood pressure sensor.

In one embodiment, the fall detection system comprises a telecommunications module for sending the alert and/or the movement data to a remote device. In one embodiment, the telecommunications module is based on one or more of the following telecommunications protocols: GSM; WiFi; and Bluetooth. In one embodiment, the telecommunications module communicates with an intermediate device which in turn communicates with the remote device.

Throughout this specification, including the claims, the words “comprise”, “comprising”, and other like terms are to be construed in an inclusive sense, that is, in the sense of “including, but not limited to”, and not in an exclusive or exhaustive sense, unless explicitly stated otherwise or the context clearly requires otherwise.

BRIEF DESCRIPTION OF THE FIGURES

Preferred embodiments in accordance with the best mode of the present invention will now be described, by way of example only, with reference to the accompanying figures, in which the same reference numerals refer to like parts throughout the figures unless otherwise specified, and in which:

Fig.1

is a schematic diagram of a fall detection system in accordance with an embodiment of the present invention;

Fig.2

is a graph of accelerometer data captured by an embodiment of the present invention during a fall event;

Fig.3

is a graph of accelerometer data captured by an embodiment of the present invention during a walk event;

Fig.4

is a flow diagram showing an algorithm employed by an embodiment of the present invention for detecting a walk event;

Fig.5

is a flow diagram showing an algorithm employed by an embodiment of the present invention for compiling a simulated timed-up-and-go (TUG) test score;

Fig.6

is a flow diagram showing an algorithm employed by an embodiment of the present invention for adjusting thresholds based on false positives;

Fig.7

is a flow diagram showing an algorithm employed by an embodiment of the present invention for adjusting thresholds based on movement parameters; and

Fig.8

is a flow diagram showing an algorithm employed by an embodiment of the present invention for adjusting multiple devices.

Referring to the figures, there is provided a fall detection system 1 comprising one or more movement sensors 2 for detecting a movement of a user. A processor 3 is provided for receiving movement data from the movement sensors 2. The processor 3 analyzes the movement data to identify a free-fall 4 by the user followed by an impact 5 by the user. The processor 3 identifies a detected fall and sends an alert signal when the free-fall reaches a free-fall threshold 6 and the impact reaches an impact threshold 7.

The processor 3 can analyze the movement data to identify a timeout period 8 beginning after the free-fall 4, with the processor 3 only sending the alert signal when the impact 5 occurs within the timeout period 8. The processor 3 can also analyze the movement data to identify a post-fall period 9 beginning after the impact 5, with the processor 3 only sending the alert signal when no further movement reaches a post-fall threshold 10 within the post-fall period 9.

Fig. 2 shows the movement data from a movement sensor 2 in the form of an accelerometer in one embodiment of the fall detection system 1. The movement data is acceleration over time, with the acceleration value shown on the y-axis against time on the x-axis. A free-fall 4 is detected in the form of a sudden downward acceleration, which is shown in Fig. 2 as a sudden dip in the acceleration value. Before the free-fall 4, the acceleration value is relatively constant, and is usually at around 1g (9.8 m/s/s) if the user is standing or walking against just the force of gravity. An impact 5 is detected after the free-fall. The impact 5 is in the form of sudden upward acceleration, shown as a sudden spike in acceleration in Fig. 2. Since the free-fall 4 reaches a free-fall threshold 6 and the impact 5 reaches an impact threshold 7, a fall is identified by the processor 3. As can be seen, the free-fall threshold 6 is a certain acceleration value below the acceleration value before the free-fall 4, and the impact threshold 7 is a certain acceleration value above the acceleration value before the free-fall 4. Fig. 2 also shows a timeout period 8 beginning after the free-fall 4. As can be seen, the impact 5 occurs within the timeout period 8 which confirms the fall. A post-fall period 9 is also analyzed after the impact 5. After the impact 5, the acceleration value returns to about the same level as before the free-fall 4. If it is an actual fall, the user does not move and the acceleration value does not vary beyond the post-fall threshold 10 which is relatively narrow band defined by a certain acceleration value closely above and closely below the acceleration value before the free-fall 4.

One or more of the free-fall threshold 6, impact threshold 7, post-fall threshold 10, timeout period 8, and post-fall period 9 is calculated and/or adjusted based on previously recorded movement data. For example, one or more movement parameters are identified from previously recorded movement data. In particular, the movement data can be analyzed to detect “points of interest” (POIs) which are snapshots of the movement data that have characteristics or features indicative of particular movement events from which specific movement parameters can be identified. One or more of the free-fall threshold 6, impact threshold 7, post-fall threshold 10, timeout period 8, and post-fall period 9 can then be calculated and/or adjusted based on one or more movement parameters identified from previously recorded movement data.

As specific examples, movement parameters can comprise one or more of the following: sit-to-stand time; walk speed; turn-around time; and stand-to-sit time. It is noted that these particular movement parameters make up the components of a timed-up-and-go (TUG) test, which is a test that is useful for assessing mobility, balance, and predicting falls, as will be described further below. Other specific movement parameters comprise one or more of the following: sit-to-stand jitter; sit-to-stand delay; walking delay; walking jitter; turn-around delay; turn-around jitter; stand-to-sit jitter; stand-to-sit delay; cadence; gait speed; step length; and step impact. A jitter is defined as shaky movement, a wobble, or any other abnormal, irregular, or unsteady movements that is indicative of a vulnerability to falling. For example, a wobble at the end of a sit-to-stand event can indicate that a person is about to faint. Shaky movement during walking can indicate a balance issue. A delay can also be indicative of a vulnerability to falling. For example, an abnormally long delay, or longer delay than that recorded for the person in the past, at the end of a sit-to-stand event can again indicate that a person is about to faint. Table 1 below lists further movement parameters utilized by embodiments of the present invention.

Fig. 7 shows an algorithm in accordance with one embodiment of the invention that is used to calculate and/or adjust thresholds and periods (such as the free-fall threshold 6, impact threshold 7, post-fall threshold 10, timeout period 8, and post-fall period 9) based on particular movement parameters of the following identified movement events: sit-to-stand, that is, when a user stands up from a sitting position; stand-to-sit, that is, when a user sits down from a standing position; and walk or turn-around. Specifically, the movement parameters used are: the rotational speed during the sit-to-stand event; the acceleration during the stand-to-sit event; and the impact resulting from steps taken by the user during the walk or turn-around event. For example, the fall detection system 1 can comprise a gyroscope sensor 2 to detect rotational speed during the sit-to-stand event. The fall detection system 1 can comprise an accelerometer sensor 2 to detect the acceleration during the stand-to-sit event and the impact resulting from steps taken by the user during the walk or turn-around event.

In Fig. 7, the first step (S1) is to identify a POI snapshot, that is, a snapshot of movement data that can be identified as a movement event, which in this case is one of sit-to-stand, stand-to-sit, and walk or turn-around. If it is sit-to-stand, the maximum gyroscope resultant value is obtained (S2a) to quantify the movement parameter of rotational speed during sit-to-stand. This value is then used to calculate a new impact threshold 7 (S3a). For example, the following equation can be used:

gyrospike_th = dvc_gyrospike_th – |(max(Gr) – dvc_gyrospike_th)| x cut_off_ratio

where

gyrospike_th is the new impact threshold 7;

dvc_gyrospike_th is the previous impact threshold 7;

max(Gr) is the maximum gyroscope resultant value; and

cut_off_ratio is a predetermined constant (e.g. 0.3).

If the movement event is stand-to-sit, the minimum accelerometer resultant value is obtained (S2b) to quantify the movement parameter of acceleration during stand-to-sit. This minimum accelerometer resultant value is then used to calculate a new free-fall threshold 6 (S3b). For example, the following equation can be used:

ff_th = dvc_ff_th – |(dvc_ff_th – min(Ar))| x (1.0 – cut_off_ratio)

where

ff _th is the new free-fall threshold 6;

dvc_ ff _th is the previous free-fall threshold 6;

min( A r ) is the minimum accelerometer resultant value; and

cut_off_ratio is a predetermined constant (e.g. 0.3).

If the movement event is walk or turn-around, the average step impact value is obtained (S2c) to quantify the movement parameter of impact resulting from steps taken by the user during the walk or turn-around event. This average step impact value is then used to calculate a new impact threshold 7 (S3c). For example, the following equation can be used:

imp_th = dvc_imp_th + |(max(Ar) – dvc_imp_th)| x cut_off_ratio

where

imp_th is the new impact threshold 7;

dvc_imp_th is the previous impact threshold 7;

max( Ar ) is the maximum accelerometer resultant value; and

cut_off_ratio is a predetermined constant (e.g. 0.3).

The new calculated threshold value is then recorded (S4) and used by the processor 3 to identify detected falls in the future.

The first step (S1) in Fig. 7 of identifying a POI snapshot can be done in many ways. For example, one algorithm in accordance with one embodiment of the invention is shown in Fig. 4. This algorithm is used to identify the movement event of a walk by a user, that is, a walk POI.

The first step (P1) is to check whether a walk POI has already been identified and acquired within a walk period. If not, the value of a walk window counter is checked (P2). If it is zero, then a reference step count is set (P3). If the walk window counter is greater than zero, then it is checked to determine whether it has reached a walk window value (P4). If the walk window value has been reached, then the walk window counter is reset (P5). If the walk window value has not yet been reached, then it is determined whether a minimum number of steps have been taken (P6). If the minimum number of steps have been taken, then a walk POI is detected (P7).

The walk POI detection algorithm above starts off with ensuring that only a single walk POI is detected in each walk period (P1). This is to prevent having too much walk data when the user is continuously walking. If there is a detected walk POI and the walk period has not yet been reset, the algorithm skips detection altogether. On the other hand, if no walk POI has been detected yet (P2), the algorithm proceeds to the next step.

A walk window counter is used to act as a timeout in detecting if a minimum number of steps has been made within a set walk window. At the start, or after the reset of the walk POI detection, the walk window counter starts off as zero. At this point, the step count is acquired and saved as reference step count (P3). The walk window counter continuously increments until it reaches the set walk window value (P5) or if a walk POI has been detected (P7), whichever comes first.

At each pass of the walk POI detection algorithm, the current step count is acquired and compared with the saved reference step count (P4). When the set minimum number of steps has been made before the walk window counter reaches the walk window value, a walk POI is detected (P7). At this point, the corresponding handling of a detected POI is carried out. This includes the acquisition of a range of sensor data, the formation of a snapshot object, and storage of this data in, for example, a flash memory 11 included with the fall detection system 1.

The POI detection algorithms, such as the above for detecting a walk POI, are performed at the same frequency that sensor data is acquired.

It is important to define what constitutes walk data, as well as the characteristics of the movement sensor data during an actual walk. In one embodiment of the fall detection system 1, the movement sensors 2 include an accelerometer gathering linear acceleration values in 3 axes (Ax, Ay, Az), a gyroscope gathering rotational velocity values in 3 axes (Gx, Gy, Gz), and a temperature sensor. An accelerometer resultant (Ar) and a gyroscope resultant (Gr) are derived from the individual axial components. The 3-axis accelerometer and 3-axis gyroscope can be in the form of a 6-axis accelerometer-gyroscope sensor chip.

In this embodiment, these movement sensors 2 are incorporated into a wearable device 12 to be worn in the right hand pocket of clothing which hugs the legs tightly, e.g. pants. It is this manner of wearing the device 12 which makes the 6-axis accelerometer-gyroscope sensor chip oriented such that: the x-axis is aligned along the vertical, with the positive end pointing upwards; the positive y-axis is pointing at the left of the device 12, when looking at it with a clip in front; the positive z-axis is pointing outward from the clip. Fig. 3 shows a graph of the sensor data, particularly of the accelerometer x-axis, gathered while a user of the device 12 is walking. This accelerometer axis is the most significant when it comes to walking as it is the axis which is orthogonal to the ground, and is normal to the direction of the pull of gravity. It should be noted that the sensor data example shown below produces negative Ax data because of the orientation of the sensors mentioned earlier. It is appreciated that the wearable device 12 can be or can be adapted to be worn at other locations on the body of the user. For example, the wearable device 12 can be worn in the left hand pocket of pants, or around the neck of the user.

A walk event is defined as when a user is making steps towards a certain distance. With this definition and the interpretation of the acquired movement sensor data, the detection of a walk POI becomes heavily dependent with the detection of steps. The difference between a walk and a step is that a walk constitutes a minimum of steps made. An algorithm for detecting a walk POI was described above.

Other embodiments of walk POI detection algorithms can be used. For example, in another embodiment, a “walk-start” event is defined as 5 steps from rest and indicates the start of a walk. In a further embodiment, a “walk” event is defined as 8 steps from the end of the “walk-start” event. In another embodiment, a “walk-turn” is defined as a 160 degree turn detected on a gyroscope during the walk event. The definitions of various events can be selected for respective groups of users, or can be adjusted over time for individual users based on previously recorded movement data.

Event	Movement Parameter	Data Components	Algorithm Overview
Walk over a known distance
	Cadence	cropped accelerometer data	acc resultant FFT-based
	step length	total walk distance, step count	total walk distance / step count
	gait speed	step length, cadence	step length * cadence
Walk over an unknown distance
	gait speed	cropped sensor data	linear acceleration calculation
	cadence	cropped accelerometer data	acc resultant FFT-based
	step length	gait speed, cadence	gait speed / cadence
	step impacts (mean, var, max, trend array)	filtered accelerometer data	g-values at the indices of the maxima corresponding to each step
	step length with respect to height	mean step lengths and height data	height / mean step length
	step impact with respect to weight	mean step impact and weight data	weight / mean step impact
Stepping
	step count (basic thresholding)	cropped accelerometer data	acc resultant thresholding
	step count (FFT-based)	cropped accelerometer data, cadence	cadence * walk duration
	step count (adjusted thresholding)	cropped accelerometer data	acc resultant thresholding
	step impact (mean, max, trend array)	cropped Gy data	Ar values at the indices of the maxima corresponding to each step
Sit-to-Stand and Stand-to-Sit
	sit-stand / stand-sit duration (mean, max)	cropped gyroscope and Ar data	get lengths of intervals in regions with sit-stand/stand-sit movement in the axis of rotation
	stand-sit Impact strength (mean, max)	cropped gyroscope and Ar data	get Ar value at indices of identified sitting motion
	resting time between repetitions (mean, max)	cropped gyroscope and Ar data	get length of intervals between identified movement regions of data
	mean angular speed of sit-stand / stand-sit	cropped gyroscope and Ar data	mean angular speed of axis of rotation
Turn-around
	turn speed (mean, max)	cropped Gy data	get average and max deg/s in identified range containing turn in Gy data
	turn angle recorded	cropped Gy data	integral of Gy in the identified range containing the turn
	turn duration	cropped Gy data	get length in time of identified range containing turn in Gy data

In Table 1, “acc” and “Ar” are abbreviations of “accelerometer”, and “Gy” is an abbreviation of “gyroscope”. “Cropped” data is when a particular window of data is analyzed. For example, for cadence, cropped accelerometer data is used. If a walk is identified over only a 10-second period in the recorded movement data, then the movement data is cropped to include only the 10-second walk period. The resultant acceleration is then converted to a frequency domain using a fast fourier transform (FFT). The frequency that is the maximum in the FFT frequency domain is then selected to calculate cadence.

As well as detecting falls, the fall detection system 1 also assesses one or more mobility or balance measures for a user and predicts falls, as briefly noted above. One test that is useful for assessing mobility and balance and predicting falls is a timed-up-and-go (TUG) test. This consists of timing a user undertaking the following actions from an initial sitting position at a starting point: stand up; walk 3 metres; turn around 180 degrees; walk 3 metres back to the starting point; and sit down again at the initial sitting position. TUG tests however require an investment of time for a test subject. These tests also need to be conducted by a health care professional in a clinical setting as part of a clinical program to be effective. This therefore involves costs for the test subject. The health care professional typically observes and manually times the TUG test during an appointment.

The present fall detection system 1 allows a simulated timed-up-and-go (TUG) test to be compiled from the movement parameters to calculate a TUG score. In particular, as described above, the fall detection system 1 can identify the following movement parameters from previously recorded movement data: sit-to-stand time; walk speed; turn-around time; and stand-to-sit time. The simulated TUG test is compiled by adding these time periods together. Advantageously, although possible, it is not necessary for these movement parameters to be identified during a contiguous movement event that consists of a sit-to-stand event, a walk event, a turn-around event, and a stand-to-sit event all performed continuously in order from one event to the next event. In other words, although a user can actually perform an actual TUG test, this is not necessary. The movement parameters that are compiled together are identified separately from previously recorded movement data as separate entities, and accordingly, can occur separately at different times and in different sequences. In fact, the user is not necessarily aware that she/he is performing a TUG test since the movement parameters are identified from movement data that is detected whilst the user is performing normal activities.

In a basic embodiment, the simulated TUG test provides a TUG score in the form of the sum of the sit-to-stand time, walk speed, turn-around time, and stand-to-sit time. This can be compared with TUG test times compiled through population studies. These comparative TUG test times can be average times, times at a particular percentile, and these can be broken down into different population groups based on age, gender, and other suitable criteria. For example, several studies indicate that a TUG test time of more than 13.5 seconds indicates a high risk of falls.

The simulated TUG test can be used to trigger clinical intervention or increased monitoring of a user. Simulated TUG tests can also be performed over time for an individual user and clinical intervention or increased monitoring can be triggered if there is an atypical increase in test time compared with previous test times for that individual.

Other fall risk scores can be calculated based on one or more of the movement parameters. For example, one fall risk score can be based on the magnitude or frequency of walk jitter. If there is an increase in magnitude and/or frequency of walk jitter, this will be reflected in the fall risk score and clinical intervention or increased monitoring can be triggered.

In general, the fall detection system 1 can comprise an analytics module for assessing the risk of a fall and/or for predicting a fall based on previously recorded movement data.

Fig. 5 shows an embodiment of the fall detection system 1 which calculates and monitors simulated TUG test scores. As well as the wearable device 12, there is also a web-based application 13, a database 14, and an analytics module 15. The web-based application receives activity logs from the device 12, calculates simulated TUG test scores (e.g. weekly, monthly), and displays simulated TUG test scores. The database 14 stores data used in the fall detection system 1, including the movement data received from the movement sensors 2, data received from the device 12, and data calculated by the fall detection system 1. The analytics module 15 identifies relevant movement parameters and calculates the components of the simulated TUG test score.

False positive data and/or true positive data can also be identified from previously recorded movement data. One or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period can then be calculated and/or adjusted based on the false positive data and/or true positive data. False positive data is the data related to a detected fall that is identified as not an actual fall, whilst true positive data is the data related to a detected fall that is identified as an actual fall. For example, in the initial stages of use of the fall detection system 1 by a particular user, the thresholds and time periods used may be simply default values or values that are not very accurate since they have not had the benefit of being set or calculated from extensive recorded movement data for the particular user. Therefore, the fall detection system 1 may not accurately detect a fall. In particular, it may detect a fall when in actual fact a fall has not occurred, that is, a false positive fall.

Fig. 6 shows an algorithm in accordance with one embodiment of the invention that is used to calculate and/or adjust thresholds and periods (such as the free-fall threshold 6, impact threshold 7, post-fall threshold 10, timeout period 8, and post-fall period 9) based on a false positive fall.

In another embodiment, the detected fall is compared with the false positive data and/or true positive data before sending the alert, and the alert is only sent when the detected fall is determined to be a true positive based on the false positive data and/or true positive data.

One or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period can also be calculated and/or adjusted based on one or more personal parameters of the user. For example, the personal parameters comprise one or more of the following: age; gender; fall history; medical history; injury history; and blood pressure. There can be default values for the thresholds or periods, or default adjustments for the thresholds or periods based on the personal parameters. For example, the post-fall period can be longer for users above a certain age to take into account slower reflexes or reaction times.

Any combination of the data used above to calculate and/or adjust thresholds and/or time periods for fall detection can also be used to assess one or more mobility or balance measures for a user and predict falls. As discussed above, fall risk scores can be calculated based on one or more of the movement parameters. In one specific example, the fall risk score (FRS) below is calculated:

FRS = (AGE_COEFF * Age) + (GENDER_COEFF * Gender) + (CADENCE_COEFF * Cadence) + (GAITSPEED_COEFF * GaitSpeed) + (STEPLENGTH_COEFF * StepLength)

where

AGE_COEFF = 0.0186;

GENDER_COEFF = -0.0452;

CADENCE_COEFF = 0.1915;

GAITSPEED_COEFF = -0.7983;

STEPLENGTH_COEFF = -0.6440; and

Gender = 0 for males, 1 for females.

The movement sensors 2 can comprise one or more of the following: an accelerometer; a 3-axis accelerometer; a 6-axis accelerometer; a 9-axis accelerometer; a multi-axis accelerometer; a gyroscope; a 3-axis gyroscope; and an altimeter. The fall detection system 1 can also comprise one or more supplemental sensors. For example, the supplemental sensors can comprise one or more of the following: a temperature sensor; a heart rate sensor; and a blood pressure sensor. These supplemental sensors can complement the movement sensors 2 and increase the accuracy of fall detection. For example, changes in heart rate and/or blood pressure typically occur during falls and these can be sensed by supplemental sensors such as a heart rate sensor and a blood pressure sensor to increase the confidence that a fall has occurred.

The fall detection system 1 can comprise a telecommunications module 16 for sending the alert and/or the movement data to a remote device 17. The telecommunications module 16 can be based on one or more of the following telecommunications protocols: GSM; WiFi; and Bluetooth. The telecommunications module 16 can communicate with an intermediate device 18 which in turn communicates with the remote device 17. The intermediate device 18 can be, for example, a mobile phone carried by a user. In this case, the telecommunications module 16 only needs to communicate using a local or short range telecommunications protocol, such as Bluetooth, since it only needs to communicate with the intermediate device 18. The intermediate device 18 then communicates with the remote device 17, including passing on communications, commands, and data from the wearable device 12 or other parts of the fall detection system 1.

As described above, the movement sensors 2 can be incorporated into a wearable device 12 worn by a user. One or more of the other parts of the fall detection system 1 such as the processor 3, the memory 11, the database 14, and the analytics module 15 can also be incorporated into the wearable device 12. However, one or more of these other parts can be located elsewhere such as the remote device 17, the intermediate device 18, or a remote server. For example, to allow the wearable device 12 to be small and lightweight, one embodiment sees the wearable device 12 incorporate the movement sensors 2, the processor 3, and the telecommunications module 16. The database 14 and the analytics module 15 are located on a remote server. The telecommunications module 16 interconnects the wearable device 12 and the remote server. In another embodiment, the wearable device 12 incorporates the movement sensors 2, the processor 3, and the telecommunications module 16. The analytics module 15 is located on an intermediate device 18, whilst the database 14 is located on a remote server. The wearable device 12 communicates with the intermediate device 18 via the telecommunications module using a short range communications protocol (e.g. Bluetooth) and communicates with the remote server through the intermediate device 18 which communicates with the remote server using a long range communications protocol (e.g. GSM). In other embodiments, instead of the wearable device 12 incorporating the processor 3, the processor 3 can be a processor incorporated in the intermediate device 18, the remote device 17, or a remote server. This allows for an especially low-cost wearable device 12. Furthermore, if an existing processor on the intermediate device 18 can be utilized as the processor 3 of the fall detection system 1 then the cost of providing a separate processor can be avoided. The wearable device 12 can still include a rudimentary processor to manage the functions of the wearable device 12 but leave the more resource hungry functions to the existing processor in the intermediate device 18.

Advantageously, having resource hungry functions on a remote server, such as the database 14 and the analytics module 15, means that a plurality of wearable devices 12 can be connected and managed by a single remote server. Respective thresholds and time periods for detecting and predicting falls for each wearable device 12 can then be calculated by the single remote server and then communicated to each wearable device 12. In this case, each wearable device 12 has a device ID so that the respective thresholds and time periods can be correctly communicated to the corresponding wearable device 12. Fig. 8 shows an embodiment of an algorithm that ensures that the respective thresholds and time periods are correctly communicated to the corresponding wearable device 12.

It is appreciated that the aforesaid embodiments are only exemplary embodiments adopted to describe the principles of the present invention, and the present invention is not merely limited thereto. Various variants and modifications can be made by those of ordinary skill in the art without departing from the spirit and essence of the present invention, and these variants and modifications are also covered within the scope of the present invention. Accordingly, although the invention has been described with reference to specific examples, it is appreciated by those skilled in the art that the invention can be embodied in many other forms. It is also appreciated by those skilled in the art that the features of the various examples described can be combined in other combinations.

Claims

A fall detection system comprising:
one or more movement sensors for detecting a movement of a user; and
a processor for receiving movement data from the movement sensors, the processor analyzing the movement data to identify a free-fall by the user followed by an impact by the user, and the processor identifying a detected fall and sending an alert signal when the free-fall reaches a free-fall threshold and the impact reaches an impact threshold.
A fall detection system according to claim 1 wherein the processor analyzes the movement data to identify a timeout period beginning after the free-fall, the processor only sending the alert signal when the impact occurs within the timeout period.
A fall detection system according to any one of claims 1 to 2 wherein the processor analyzes the movement data to identify a post-fall period beginning after the impact, the processor only sending the alert signal when no further movement reaches a post-fall threshold within the post-fall period.
A fall detection system according to any one of claims 1 to 3 wherein one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on previously recorded movement data.
A fall detection system according to any one of claims 1 to 3 wherein one or more movement parameters are identified from previously recorded movement data.
A fall detection system according to claim 5 wherein the movement parameters comprise one or more of the following: sit-to-stand time; walk speed; turn-around time; and stand-to-sit time.
A fall detection system according to claim 6 wherein a simulated timed-up-and-go (TUG) test is compiled from the movement parameters to calculate a TUG score.
A fall detection system according to claim 5 wherein the movement parameters comprise one or more of the following: sit-to-stand jitter; sit-to-stand delay; walking delay; walking jitter; turn-around delay; turn-around jitter; stand-to-sit jitter; stand-to-sit delay; cadence; gait speed; step length; and step impact.
A fall detection system according to any one claims 5 to 8 wherein a fall risk score is calculated based on one or more of the movement parameters.
A fall detection system according to any one of claims 5 to 9 wherein one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on one or more movement parameters identified from previously recorded movement data.
A fall detection system according to any one of claims 1 to 10 wherein false positive data and/or true positive data are identified from previously recorded movement data.
A fall detection system according to claim 11 wherein one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on the false positive data and/or true positive data.
A fall detection system according to any one of claims 11 to 12 wherein the detected fall is compared with the false positive data and/or true positive data before sending the alert, and the alert is only sent when the detected fall is determined to be a true positive based on the false positive data and/or true positive data.
A fall detection system according to any one of claims 1 to 13 wherein one or more of the free-fall threshold, impact threshold, post-fall threshold, timeout period, and post-fall period is calculated and/or adjusted based on one or more personal parameters of the user.
A fall detection system according to claim 14 wherein the personal parameters comprise one or more of the following: age; gender; fall history; medical history; injury history; and blood pressure.
A fall detection system according to any one of claims 1 to 15 comprising an analytics module for assessing the risk of a fall and/or for predicting a fall based on previously recorded movement data.
A fall detection system according to any one of claims 1 to 16 wherein the movement sensors comprise one or more of the following: an accelerometer; a 3-axis accelerometer; a 6-axis accelerometer; a 9-axis accelerometer; a multi-axis accelerometer; a gyroscope; a 3-axis gyroscope; and an altimeter.
A fall detection system according to any one of claims 1 to 17 comprising one or more supplemental sensors.
A fall detection system according to claim 18 wherein the supplemental sensors comprise one or more of the following: a temperature sensor; a heart rate sensor; and a blood pressure sensor.
A fall detection system according to any one of claims 1 to 19 comprising a telecommunications module for sending the alert and/or the movement data to a remote device.
A fall detection system according to claim 20 wherein the telecommunications module is based on one or more of the following telecommunications protocols: GSM; WiFi; and Bluetooth.
A fall detection system according to any one of claims 20 to 21 wherein the telecommunications module communicates with an intermediate device which in turn communicates with the remote device.