AU2021103899B4

AU2021103899B4 - Radar-based electromagnetic wave fall detection system

Info

Publication number: AU2021103899B4
Application number: AU2021103899A
Authority: AU
Inventors: Changyang Li; Zhiyu Ning
Original assignee: Rudder Technology Pty Ltd
Current assignee: Rudder Technology Pty Ltd
Priority date: 2021-07-06
Filing date: 2021-07-06
Publication date: 2021-11-11
Anticipated expiration: 2029-07-06
Also published as: AU2021103899A4

Abstract

A radar-based electromagnetic wave fall detection system comprises a detector in operable communication with a radar unit and a floor vibration sensor. The detector is configured to detect a fall using signals obtained from both the radar unit and the vibration sensor, thereby improving detection accuracy of radar-based fall detection systems. 8

Description

Radar-based electromagnetic wave fall detection system

Field of the Invention

[0001] This invention relates generally to fall detection system and, more particularly, this invention relates to a radar-based electromagnetic wave fall detection system.

Background of the Invention

[0002] Falls are one of the leading causes of injury and death of the elderly and, as such, round-the-clock fall detection systems can form an integral part of aged care.

[0003] Conventional wearable devices (such as smart watches) have limitations, including requiring frequent charging. Furthermore, video surveillance cannot be used in bedrooms or bathrooms for privacy reasons.

[0004] More recently, radar-based electromagnetic wave fall detection systems show promise for in-home monitoring of the elderly for being non-obstructive, nonintrusive and insensitive to lighting and which are therefore able to preserve privacy and safety.

[0005] However, whereas radar-based systems may achieve satisfactory detection rates in experimental laboratory environments, detection rates could be improved for real-world environments.

[0006] The present invention seeks to provide a radar-based electromagnetic wave fall detection systems which provides more accurate fall detection in real-world environments, which will overcome or substantially ameliorate at least some of the deficiencies of the prior art, or to at least provide an alternative.

[0007] It is to be understood that, if any prior art information is referred to herein, such reference does not constitute an admission that the information forms part of the common general knowledge in the art, in Australia or any other country.

Summary of the Disclosure

[0008] There is provided herein a radar-based electromagnetic wave fall detection system comprises a detector in operable communication with a radar unit and a floor vibration sensor. The detector is configured to detect a fall using signals obtained from both the radar unit and the vibration sensor, thereby improving detection accuracy of radar-based fall detection systems.

[0009] The detector may comprise an optimised feature extractor and classifier which

classifies frequency domain features of signals obtained from the radar unit into fall

and non-full classifications.

[0010] The detector may also resolve a location of a person using signals obtained

from the radar unit.

[0011] Furthermore, the detector may determine if amplitude of signals obtained from

the vibration sensor exceeds a threshold.

[0012] The detector may detector a fall using a combination of the fall or non-fall

classifications, location and whether the amplitude exceed the threshold.

[0013] Other aspects of the invention are also disclosed.

Brief Description of the Drawings

[0014] Notwithstanding any other forms which may fall within the scope of the present

invention, preferred embodiments of the disclosure will now be described, by way of

example only, with reference to the accompanying drawings in which:

[0015] Figure 1 shows an exemplary radar-based fall detection system in accordance

with an embodiment; and

[0016] Figure 2 shows exemplary data processing of the system of Figure 1 in

accordance with an embodiment.

Description of Embodiments

[0017] Figure 1 shows a fall detector 101 for a home environment, aged care facility

or the like.

[0018] The detector 101 comprises a processor 106 for processing digital data. A

memory device 102 configured for storing digital data including computer program

code instructions is operably coupled to the processor 106 via a system bus 107. In

use, the processor fetches computer program code instructions and associated data

104 from the memory device 102 for interpretation and execution of the computational

functionality provided herein.

[0019] The detector 101 comprises an 1/O interface 105 in operable communication

with a radar unit 108 and a floor vibration sensor 112. In the embodiment shown, the radar unit 108 is installed within a ceiling 109 and the floor vibration sensor 112 is operably coupled to a floor surface 111 beneath the ceiling 109. However, it should be appreciated that the radar unit 108 and vibration sensor 112 may be installed in other locations.

[0020] The detector may be configured to detect a fall of a person 113 within a

supervision area 110.

[0021] The radar unit 108 may be an ultra-wideband (UWB) radar unit. The radar unit

108 may detect Doppler effect radar backscatter signals from a moving object/person

113 which are interpreted by the detector 101. The Doppler effect radar backscatter

signals may be used to infer a change in height (AZ) of the person 113. Signals from

the radar unit 108 may be further used to infer X and Y coordinates so that detection

may be confined to the supervision area 110.

[0022] The vibration sensor 112 may be a solid-state vibration sensor 112 operably

coupled to a floor surface to detect vibrations transmitted therethrough when the

person 113 hits the floor 111. The vibration sensor 112 may be configured to measure

vibration amplitude and/or frequency.

[0023] The detector 101 comprises an 1/O interface 105 operably interfacing the radar

unit 108 and the floor vibration sensor 112 which may digitise signals obtained from

the radar unit 108 and vibration sensor 112.

[0024] The computer program code instructions may be logically divided into a

plurality of computer program code instruction controllers 103. As will be described

in further detail below, the controllers 103 may comprise a Fast Fourier Transform

controller which converts signals from the radar unit 108 into a frequency distribution

117.

[0025] An optimised feature extractor may extract dominant or pertinent frequency

domain features from the frequency distribution 117 and a classifier 119 (which may

implement a softmax/normalized exponential function) may normalize the output of

the feature extractor 118 to a probability distribution over predicted output classes,

the output classes in this case being fall and non-fall classes 124.

[0026] The controllers may further resolve the signals obtained from the radial unit

into x, y and/or z coordinates to determine a location 125 of the moving object/person.

The controllers may determine whether the location falls within the supervision area

110.

[0027] The controllers may further determine if the amplitude 121 (and, in

embodiments, frequency) of the signals from the vibration sensor 112 exceed a

threshold.

[0028] The controllers may further comprise a combination function controller which

implements a combination function 122 using the classification 124, location 124 and

if the amplitude 121 exceeds the threshold to make a decision 123 indicative of

whether a fall is detected.

[0029] If a fall is detected, the detector 101 may transmit a signal via the 1/O interface

101 to an alarm unit, via a communication system or the like.

[0030] Fall detection by the detector 101 will now be described in further detail with

reference to Figure 2 in accordance with an exemplary embodiment.

[0031] The detector 101 may transform radar signal signals from the radar unit 101

into the frequency distribution 117, such as using a Fast Fourier Transform.

[0032] Then, the detector 101 employs the feature extractor 118 to extract dominant

and/or pertinent frequency domain features from the frequency distribution 117.

[0033] The feature extractor 118 may employ machine learning wherein the feature

extractor 118 is a random forest, support vector machine or neural network optimised

to detect falls using signals from the radar unit 108.

[0034] In accordance with one embodiment, the feature extractor 118 comprises

several convolution layers and to fully connected layers as shown in the following

table:

LayerIndex Filter Size Filter Stride Pad Filter Channel 1 7x7 3 0 128 2 5x5 3 0 128 3 3x3 1 1 128 4 3x3 1 1 256 5 3x3 1 1 512 6 (fully - - - 1024 connected) 7 (fully -- 2 connected)

[0035] A softmax function may be applied by the classifier 124 to the features

extracted by the feature extractor 118 to make the classification 124.

[0036] The following loss function may be used to train the feature extractor 118:

[0037] [Equation 1] Li = I * log(p) - (1 - 1) * log (1 - p),

[0038] where I = 1 if fall happens and 0 otherwise; p is the output of the feature

extractor.

[0039] Signals from the radar unit 108 may also be interpreted to determine if the

location 125 is within the supervision area 110 according to the following equation:

[0040] [Equation 2] D2 = ((x and y coordinateare in supervision area) x ((Az),

[0041]where ((-) is 1 if the condition in the parenthesis is true and 0 otherwise; Az

represent that the z coordinate decreases dcm in t second period.

[0042] Additionally, output from the vibration sensor 116 may be monitored according

to the following detection function:

[0043] [Equation 3] D3 = ((v),

[0044] where v is the vibration intensity change in t second period.

[0045] Equations 1 - 3 may be combined using a combination function 122 to detect

falls.

[0046] In accordance with a first embodiment, the combination function 122 is an

additive combination function wherein:

[0047] [Equation 4] J1 = a x p + l x D2 + y x D3,

[0048] where a,#, and y control the power of Equation 1-3, respectively.

[0049] Alternatively, the combination function 122 is a multiplicative function as

follows:

[0050] [Equation 5] J2 = p x D2 x D3

[0051] The decision 123 may be determined according to the following function:

[0052] [Equation 6] J1 or J2 > Threshold

[0053] The foregoing description, for purposes of explanation, used specific

nomenclature to provide a thorough understanding of the invention. However, it will

be apparent to one skilled in the art that specific details are not required in order to

practise the invention. Thus, the foregoing descriptions of specific embodiments of

the invention are presented for purposes of illustration and description. They are not

intended to be exhaustive or to limit the invention to the precise forms disclosed as

obviously many modifications and variations are possible in view of the above

teachings. The embodiments were chosen and described in order to best explain the

principles of the invention and its practical applications, thereby enabling others

skilled in the art to best utilize the invention and various embodiments with various

modifications as are suited to the particular use contemplated. It is intended that the

following claims and their equivalents define the scope of the invention.

[0054] The term "approximately" or similar as used herein should be construed as

being within 10% of the value stated unless otherwise indicated.

Claims

Claims 1. A fall detection system comprising a detector in operable communication with

a radar unit and a floor vibration sensor and wherein the detector is configured to

detect a fall using signals obtained from the radar unit and the vibration sensor,

wherein the detector comprises an optimised feature extractor and classifier which

classifies frequency domain features of signals obtained from the radar unit into fall

or non-fall classifications.
2. The system as claimed in claim 1, wherein the detector resolves a location of a

person using the signals obtained from the radar unit.
3. The system as claimed in claim 1, wherein the detector determines if

amplitude of signals obtained from the vibration sensor exceed a threshold.
4. The system as claimed in claims 1 - 3, wherein the fall detection system

detects a fall using a combination of the fall or non-fall classifications, location and

whether the amplitude exceed the threshold.