Healthcare-Oriented Characterisation of Human Movements

by Means of Impulse-Radar Sensors

and by Means of Accelerometric Sensors

Paweł Mazurek

, Jakub Wagner

, Andrzej Miękina

, Roman Z. Morawski

and Frode Fadnes Jacobsen

Institute of Radioelectronics and Multimedia Technology, Faculty of Electronics and Information Technology,

Warsaw University of Technology, Nowowiejska 15/19, 00-665 Warsaw, Poland

Bergen University College, Center for Care Research, Møllendalsveien 6-8, 5020 Bergen, Norway

Keywords: Healthcare, Impulse-Radar Sensor, Accelerometer, Measurement Data Processing, Uncertainty Estimation.

Abstract: This paper is devoted to the healthcare-oriented characterisation of the human movements by means of the

accelerometric and impulse-radar sensors – the sensors that may be employed in care services for

monitoring of elderly and disabled persons. Characterisation of the movements in terms of the so-called

self-selected walking velocity can be used by the medical and healthcare personnel to assess the overall

health status of a monitored person. The quality of the characterisation, based on the measurement data from

accelerometric and impulse-radar sensors, has been assessed in a series of real-world experiments which

involved the estimation of the instantaneous and mean walking velocity of a person moving according to

predefined patterns. Some indicators of uncertainty of the velocity estimation, determined with respect to

assumed predefined velocity values, have been used for comparison of the performance of both types of

sensors. The experiments have shown that impulse-radar sensors enable one to estimate the mean walking

velocity more accurately than the accelerometric sensors: the estimates obtained on the basis of data from

the latter sensors are affected by larger bias and are more widely spread around their mean values.

1 INTRODUCTION

The life expectancy has been growing in Europe for

many years, while the healthy life expectancy has

been slightly diminishing since the last decade of the

XXth century (cf. http://www.healthy-life-years.eu/).

Hence the growing importance of research on new

technologies that could be employed in monitoring

systems supporting care services for elderly and

disabled persons. The capability of those systems to

detect dangerous events, such as person’s fall, is of

key importance (Hamm et al., 2016). However,

those systems are expected not only to detect

dangerous events, but also to predict those events on

the basis of acquired data. The analysis of gait, as

well as of the itinerary and timing of activities of the

monitored persons, may thus contribute to the

prevention (Baldewijns et al., 2016a). The relevance

of features related to gait analysis in monitoring of

elderly persons, and in particular – in fall

prevention, has been emphasised in several recent

papers (Buracchio et al., 2010, Studenski et al.,

2011, Lusardi, 2012, Egerton et al., 2014, Stone et

al., 2015, Thingstad et al., 2015, Baldewijns et al.,

2016b).

So far, the most popular monitoring technique,

already applied in healthcare practice, is based on

wearable devices (Bulling et al., 2014, Cola et al.,

2014, Luque et al., 2014, Brodie et al., 2015). Those

devices do not require a pre-built infrastructure and

thus may be used outdoor. The signals from

movement sensors (mainly accelerometers and

gyroscopes), worn by a monitored person, are

transmitted via radio links to a computer and

analysed. This solution makes also possible the

acquisition of physiological data (such as values of

blood pressure, ECG data or EEG data).

Recently, numerous attempts have been made to

apply various radar techniques for monitoring of

elderly and disabled persons (Cuddihy et al., 2012,

Liu et al., 2012, Tomii and Ohtsuki, 2012, Jian et al.,

2014, Su et al., 2015, Miękina et al., 2016b). Those

attempts are mainly motivated by the conviction that

128

Mazurek P., Wagner J., MiÄ

Zkina A., Morawski R. and Jacobsen F.

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors.

DOI: 10.5220/0006154201280138

In Proceedings of the 10th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2017), pages 128-138

ISBN: 978-989-758-213-4

those techniques may be less intrusive than vision-

based solutions (e.g. digital cameras), less

cumbersome than wearable solutions (e.g.

accelerometers and gyroscopes), and less invasive

with respect to the home environment than

environmental solutions (e.g. pressure sensors).

This paper is devoted to the assessment of the

uncertainty of the estimation of the walking velocity,

on the basis of data acquired by means of impulse-

radar sensors and by means of accelerometric

sensors. As suggested in the literature, e.g. (Fritz and

Lusardi, 2009), the walking velocity is highly

informative for healthcare experts; for example:

 the velocity lower than 0.6 m/s enables them to

predict an increase in the risk of falls and

hospitalisation of a monitored person;

 an improvement in walking velocity of at least

0.1 m/s is a useful predictor for well-being;

 a decrease of the same amount is correlated with

deterioration of the health status or advancement

of disability.

The comparative study, reported in this paper, is

based on an extensive set of real-world experiments

which comprise:

 simultaneous recording of measurement data

from both types of sensors, representative of the

gait characteristics of a person moving according

to predefined patterns;

 statistical analysis of those data, aimed at

determination of certain indicators of uncertainty

of the velocity estimation.

Due to the operation principle of both types of

sensors, one may expect that the position of a

monitored person can be better estimated on the

basis of the data from impulse-radar sensors

(hereinafter called radar data for brevity), and its

acceleration – on the basis of data from the

accelerometric sensors (hereinafter called

accelerometric data for brevity). Therefore, despite

the fact that both the position and the acceleration

may also be of interest for the healthcare personnel,

this study is confined to the uncertainty of the

estimation of the velocity, which requires similar

degree of the measurement data preprocessing for

both types of sensors.

2 METHODOLOGY OF

EXPERIMENTATION

2.1 Data Acquisition

The raw measurement data for experimentation have

been acquired by means of the APDM Opal

accelerometric sensor (cf. http://www.apdm.com/

wearable-sensors/) attached to the waist of a

monitored person, and by means of a pair of

synchronised impulse-radar sensors – cf. (Morawski

et al., 2014) – whose location is shown in figure 1. A

monitored person has moved at the distance of ca.

1–6.5 m from each of them.

The walking velocity has been assessed on the

basis of real-world data acquired when an

experimenter has been walking at various constant

velocities, ranging from 0.5 m/s to 1.0 m/s, forth and

back along a straight line –

20R 

times along the

x-axis, between points (0,3) and (4,3), and

times

along the y-axis, between points (2,1) and (2,5) (cf.

figure 1). In order to assure a known constant

walking velocity, a metronome has been used.

Figure 1: Experimental setup; the crosses indicate the

reference points, i.e. the points where marks have been

placed on the floor.

2.2 Data Preprocessing

2.2.1 Radar Data

The measurement data from a pair of impulse-radar

sensors – after preliminary preprocessing, as

described in (Miękina et al., 2016a) – take on the

form of a sequence of numbers representative of the

x-y coordinates of a monitored person.

A sequence of the estimates of the instantaneous

walking velocity may be obtained by numerical

differentiation of the sequence of the position

estimates, e.g. by means of the central-difference

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors

129

method (Wagner et al., 2015), defined by the

formula:

(1)

for 1, , 1

dnN









 (1)

where





is a sequence of data to be

differentiated, and

11nn n

tt t



 

are the

differentiation steps, with

denoting the time

moments at which the data have been acquired. That

method is, however, very sensitive to errors

corrupting the data used for derivative estimation;

therefore, it should be regularised through, e.g.,

optimisation of the differentiation step. The total

velocity magnitude has been calculated according to

the formula:



(1) (1)

ˆˆ ˆ

for 1, ,

nn n

vx y n N 

(2)

where

(1)

and

(1)

are estimates of the first

derivatives, computed on the basis of the estimates

of the x- and y-data sequences.

2.2.2 Accelerometric Data

An accelerometric sensor – composed of an

accelerometer, magnetometer and gyroscope –

provides a sequence of data representative of the

monitored person’s instantaneous acceleration in

three directions, viz. magnetic north, magnetic west,

and vertical. A sequence of the estimates of the

instantaneous velocities in these directions can be

obtained by numerical integration of the sequences

of the acceleration values. It must be, however,

taken into account that – since both systematic and

random errors corrupting accelerometric data

propagate through the integration process (Thong et

al., 2004) – the velocity estimates may be subject to

a growing-with-time drift and random errors whose

standard deviation is also growing with time. As a

consequence, non-zero estimates may appear even

when a monitored person is standing still; therefore,

the velocity estimates have to be corrected by means

of a so-called zero-velocity compensation procedure

(Bang et al., 2003). It can be applied to a velocity

trajectory whose first and last values are known to

be zero. In the research reported here, the following

correction formula has been used:

ˆˆ









for

nnn

(3)

where

ˆˆ





and

are the indices of the

first and last time instants of the movement,

respectively; the latter parameters have been

determined experimentally. The corrected velocity

trajectories in the magnetic north and west directions

(denoted with

and

, respectively) have been

used for computing the total velocity magnitude

according to the formula:



ˆˆ ˆ

nn n

vv v

for

1, ,nN 

(4)

2.3 Criteria of Performance Evaluation

In each experiment,

sequences of the

instantaneous total velocity estimates have been

computed using equations 2 and 4 on the basis of

both radar data and accelerometric data:





()

| 1,...,

vn N

for

1, ,rR 

(5)

Prior to the evaluation of the uncertainty of the

estimation, some outlying sequences have been

removed to prevent the misinterpretation of the

results. The outlying sequences have been identified

as those whose mean value:

() ()

ˆˆ







(6)

deviated from the group mean value:

()

ˆˆ







(7)

by more than three standard deviations:



()

ˆˆˆ











(8)

Next, the qualitative assessment of the

uncertainty of the estimates has been performed. It

has been based on the inspection of the estimates of

the mean:

()

ˆˆ











(9)

and standard deviation:



()

ˆˆˆ

nnn















(10)

of each element of the sequence of the instantaneous

velocity estimates;



denotes the number of

sequences in a set under consideration after

removing the outlying sequences.

Finally, the quantitative assessment of the

uncertainty of the estimates of the mean walking

velocity has been done using the following

indicators:

 the absolute discrepancy between the mean value

of the estimates of the velocity and the

HEALTHINF 2017 - 10th International Conference on Health Informatics

130

predefined value of that velocity,

 the absolute root-mean-square discrepancy of the

estimates with respect to the predefined value,

 the lower and upper bounds of the absolute

discrepancy between the estimates and the

predefined value.

The above indicators have been calculated

separately for each set of



estimates of mean

walking velocity, obtained in each experiment by

averaging its

samples.

3 RESULTS AND DISCUSSION

In figures 2–5, the mean sequences of instantaneous

velocity estimates of a moving person, obtained on

the basis of the radar data and accelerometric data –

for both directions of movement (i.e. along x-axis

and along y-axis) and for all predefined velocity

values – are presented.

It is worth being noticed that the uncertainty of

estimation, based on radar data, is direction

dependent: for the movement along the x-axis and

predefined velocity values from 0.5 m/s to 0.7 m/s,

the estimated mean value of the velocity oscillates

around the predefined value during the movement.

This cannot be observed for the movement along the

y-axis in the same range of velocity values.

Moreover, it may be seen that the standard deviation

of the velocity is greater for the movement along the

x-axis.

Those differences are caused by the fact, that the

calculation of the position of the moving person is

easier when the distance between the person and

each of the radars is equal (i.e. when each radar sees

the same side of the human body).

On the other hand, it may be noticed that the

uncertainty of estimation, based on accelerometric

data, is direction independent.

a) b)

Figure 2: Uncertainty indicators determined for estimates of the velocity of a moving person, obtained on the basis of the

radar data (a) and accelerometric data (b), for the movement along x-axis with the velocity values ranging from 0.5 m/s to

0.6 m/s. In all sub-figures: the thick solid line denotes the sequence of mean values, while the dotted lines – the sequences

of values that are three standard deviations away from the mean sequence.

456789101112

Time [s]

-0.5

0.5

1.5

Predefined velocity = 0.5 m/s

Predefined value

Mean value

Mean value +/- 3std

345678910

Time

[

]

-0.5

0.5

1.5

Predefined velocity = 0.6 m/s

Predefined value

Mean value

Mean value +/- 3std

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors

131

a) b)

Figure 3: Uncertainty indicators determined for estimates of the velocity of a moving person, obtained on the basis of the

radar data (a) and accelerometric data (b), for the movement along x-axis with the velocity values ranging from 0.7 m/s to

1.0 m/s. In all sub-figures: the thick solid line denotes the sequence of mean values, while the dotted lines – the sequences

of values that are three standard deviations away from the mean sequence.

2345678

Time

[

]

-0.5

0.5

1.5

Predefined velocity = 0.8 m/s

Predefined value

Mean value

Mean value +/- 3std

23456

Time [s]

-0.5

0.5

1.5

Predefined velocity = 1 m/s

Predefined value

Mean value

Mean value +/- 3std

23456

Time

[

]

-0.5

0.5

1.5

Predefined velocity = 1 m/s

Predefined value

Mean value

Mean value +/- 3std

HEALTHINF 2017 - 10th International Conference on Health Informatics

132

a) b)

Figure 4: Uncertainty indicators determined for estimates of the velocity of a moving person, obtained on the basis of the

radar data (a) and accelerometric data (b), for the movement along y-axis with the velocity values ranging from 0.5 m/s to

0.8 m/s. In all sub-figures: the thick solid line denotes the sequence of mean values, while the dotted lines – the sequences

of values that are three standard deviations away from the mean sequence.

456789101112

Time [s

]

-0.5

0.5

1.5

Predefined velocity = 0.5 m/s

Predefined value

Mean value

Mean value +/- 3std

345678910

Time [s

]

-0.5

0.5

1.5

Predefined velocity = 0.6 m/s

Predefined value

Mean value

Mean value +/- 3std

3456789

Time [s

]

-0.5

0.5

1.5

Predefined velocity = 0.7 m/s

Predefined value

Mean value

Mean value +/- 3std

2345678

Time [s]

-0.5

0.5

1.5

Predefined velocity = 0.8 m/s

Predefined value

Mean value

Mean value +/- 3std

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors

133

a) b)

Figure 5: Uncertainty indicators determined for estimates of the velocity of a moving person, obtained on the basis of the

radar data (a) and accelerometric data (b), for the movement along y-axis with the velocity values ranging from 0.9 m/s to

1.0 m/s. In all sub-figures: the thick solid line denotes the sequence of mean values, while the dotted lines – the sequences

of values that are three standard deviations away from the mean sequence.

In figure 6, the so-called box plots representing

the aggregated uncertainty of the estimation of the

mean walking velocity, performed on the basis of

the radar data and accelerometric data, for each

investigated value of the walking velocity, are

presented. Each box plot indicates:

 the median value;

 the interquartile range (IQR), i.e. range between

the first and third quartile;

 the smallest value still within 1.5 IQR from the

first quartile, and the largest value still within 1.5

IQR from the third quartile;

 the values lying outside 1.5 IQR from the first

quartile and 1.5 IQR from the third quartile

(marked with crosses).

In table 1 and table 2, the numerical results of all

experiments – performed for various walking

velocities – are collected.

The results presented in tables 1 and 2 show that

the estimates of the mean walking velocity, obtained

on the basis of the radar data, are far more accurate

than those obtained on the basis of the

accelerometric data. For the estimation of the

velocity based on the radar data the mean

discrepancy, i.e. the difference between estimated

mean value and a predefined value of the velocity,

varies from –0.12 to 0.03 m/s, while it varies from

–0.18 to 0.24 m/s for the estimation based on

accelerometric data. Moreover, it can be observed

that the estimates obtained on the basis of the radar

data are more concentrated around their mean values

– the root-mean-square discrepancy of the radar-

data-based velocity estimates varies from 0.02 to

0.12 m/s, while it varies from 0.08 to 0.27 m/s for

the accelerometric-data-based estimates.

It can also be noticed that the estimates of the

mean walking velocity, obtained on the basis of the

radar data, tend to be underrated with respect to the

predefined walking velocity for the movements

along the x-axis, and very accurate for the

234567

Time [s

]

-0.5

0.5

1.5

Predefined velocity = 0.9 m/s

Predefined value

Mean value

Mean value +/- 3std

23456

Time

[

]

-0.5

0.5

1.5

Predefined velocity = 1 m/s

Predefined value

Mean value

Mean value +/- 3std

HEALTHINF 2017 - 10th International Conference on Health Informatics

134

Figure 6: Box plots representing the uncertainty of the estimation of the walking velocity, based on the radar data (R-

estimates) and accelerometric data (A-estimates); a) x-axis movement, b) y-axis movement.

movements along the y-axis. On the other hand, the

estimates of the mean walking velocity, obtained on

the basis of the accelerometric data, seem to be

underrated for lower walking velocities and

overrated for faster movements.

Lastly, it should be noted that the impact of the

imperfections of the movements of the experimenter,

reproducing the predefined patterns, are the same for

both sensors; so, not changing the result of

comparison.

4 CONCLUSIONS

The novelty of the study, whose results are presented

in this paper, consists in systematic comparison of

two monitoring techniques, viz. impulse-radar

sensors and accelerometric sensors, when applied for

healthcare-oriented characterisation of the human

movements.

The performance of both types of sensors has

been compared on the basis of data acquired by

means of them in a series of real-world experiments

which involved tracking of a person moving

according to predefined patterns. The indicators of

uncertainty of the velocity estimation have been

determined with respect to the assumed predefined

values of velocity.

Prior to the evaluation of the uncertainty, the

measurement data from both types of sensors have

to be adequately processed. The velocity estimates,

obtained on the basis of the accelerometric data, are

determined by numerical integration of the

sequences of the acceleration estimates and

corrected by means of a zero-velocity compensation

procedure. The velocity estimates, obtained on the

basis of the radar data, are determined using the

regularised numerical differentiation of the sequence

of the position estimates.

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors

135

Table 1: Uncertainty of mean velocity estimation for the movement along x-axis.

Uncertainty indicators characterising

estimates of mean velocity

Predefined walking velocity [m/s]

0.50 0.60 0.70 0.80 0.90 1.00

Impulse-radar sensors

Mean discrepancy [m/s] 0.03 0.04 0.06 0.08 0.08 0.12

Root-mean-square discrepancy [m/s] 0.04 0.05 0.06 0.08 0.09 0.12

Upper bound of the discrepancy [m/s] 0.01  0.03 0.03 0.06 0.07

Lower bound of the discrepancy [m/s] 0.05 0.07 0.09 0.13 0.12 0.14

Accelerometric sensors

Mean discrepancy [m/s] 0.17 0.06 0.12 0.12 0.18 

Root-mean-square discrepancy [m/s] 0.19 0.17 0.17 0.14 0.23 0.13

Upper bound of the discrepancy [m/s] 0.01 0.27 0.42 0.25 0.44 0.29

Lower bound of the discrepancy [m/s] 0.29 0.41 0.18 0.03 0.04 0.18

Table 2: Uncertainty of mean velocity estimation for the movement along y-axis.

Uncertainty indicators characterising

estimates of mean velocity

Predefined walking velocity [m/s]

0.50 0.60 0.70 0.80 0.90 1.00

Impulse-radar sensors

Mean discrepancy[m/s] 0.02 0.03 0.03   

Root-mean-square discrepancy [m/s] 0.03 0.03 0.03 0.03 0.03 0.02

Upper bound of the discrepancy [m/s] 0.04 0.06 0.07 0.05 0.07 0.04

Lower bound of the discrepancy [m/s] 0.02 0.01 0.01 0.01 0.04 0.03

Accelerometric sensors

Mean discrepancy [m/s] 0.18  0.24 0.12 0.17 0.07

Root-mean-square discrepancy [m/s] 0.19 0.21 0.27 0.13 0.18 0.08

Upper bound of the discrepancy [m/s] 0.06 0.55 0.43 0.21 0.29 0.18

Lower bound of the discrepancy [m/s] 0.27 0.40 0.01 0.03 0.06 0.03

The experiments performed have demonstrated

that impulse-radar sensors enable one to estimate the

walking velocity more accurately than the

accelerometric sensors. The estimates obtained on

the basis of data from the latter sensors are affected

by larger bias and are more widely spread around

their mean values.

Since falls among elderly persons are the main

cause of their admission and long-term stay in

hospitals (Abbate et al., 2010), the systems for

monitoring of elderly and disabled persons are

expected to perform some functions related to fall

prevention and/or fall detection. The functions

related to fall prevention are implemented to

overcome fall risk factors, implied by natural aging-

related physical disabilities, and promptly indicate

the increasing risk of falling; the functions related to

fall detection are to reliably detect falls, when they

occur, and minimise the potential injuries. Sensors

used for fall prevention are expected to be accurate

enough to enable the monitoring system to identify

changes in the monitored person’s health status on

the basis of relatively slow and subtle changes in

his/her gait characteristics, e.g. changes of the mean

HEALTHINF 2017 - 10th International Conference on Health Informatics

136

walking velocity. Sensors used for fall detection

should be selected and optimised with respect to

their sensitivity as to enable the monitoring system

to detect short abrupt changes in person’s velocity or

acceleration.

In light of the results presented in this paper, the

impulse-radar sensors seem to be promising means

for reliable fall prevention since they enable the

through-the-wall monitoring of persons (as the

electromagnetic waves propagate through non-metal

objects) and highly accurate estimation of their

velocity; those sensors are, however, less

appropriate for fall detection because of the

relatively low rate of data acquisition. On the other

hand, the accelerometric sensors appear to be not

well-suited for the long-term monitoring of the

person’s gait characteristics, but better satisfy the

requirements related to fall detection, due to their

higher sensitivity, significantly higher rate of data

acquisition, and suitability for outdoor use.

One may thus conclude that both types of sensors

studied in this paper, viz. impulse-radar sensors and

accelerometric sensors, are in some way

complementary, and therefore the combined use of

both of them may contribute to the increase in the

reliability of the monitoring of elderly and disabled

persons.

ACKNOWLEDGEMENTS

This work has been initiated within the project

PL12-0001 financially supported by EEA Grants –

Norway Grants (http://eeagrants.org/project-portal/

project/PL12-0001), and finished within the

statutory project supported by the Institute of

Radioelectronics and Multimedia Technology,

Faculty of Electronics and Information Technology,

Warsaw University of Technology.

REFERENCES

Abbate, S., Avvenuti, M., Corsini, P., Light, J. and

Vecchio, A. 2010. Monitoring of Human Movements

for Fall Detection and Activities Recognition in

Elderly Care Using Wireless Sensor Network: a

Survey. In Merrett, G. V. & Yen Kheng Tan (eds.)

Wireless Sensor Networks: Application – Centric

Design. Intech.

Baldewijns, G., Debard, G., Van Den Broeck, B., Mertens,

M., Karsmakers, P., Croonenborghs, T. and

Vanrumste, B. 2016a. Fall prevention and detection. In

Florez-Revuelta, F. & Chaaraoui, A. A. (eds.) Active

and Assisted Living: Technologies and Applications.

Herts, UK: IET.

Baldewijns, G., Luca, S., Vanrumste, B. and

Croonenborghs, T., 2016b. Developing a system that

can automatically detect health changes using transfer

times of older adults. BMC Medical Research

Methodology, vol., pp. 16-23.

Bang, W.-C., Chang, W., Kang, K.-H., Choi, E.-S.,

Potanin, A. and Kim, D.-Y., 2003. Self-contained

Spatial Input Device for Wearable Computers. In

Proc. 7th IEEE International Symposium on Wearable

Computers (White Plains, NY, USA), pp. 26-34.

Brodie, M. A., Lord, S. R., Coppens, M. J., Annegarn, J.

and Delbaere, K., 2015. Eight-Week Remote

Monitoring Using a Freely Worn Device Reveals

Unstable Gait Patterns in Older Fallers. IEEE

Transactions on Biomedical Engineering, vol. 62, pp.

2588-2594.

Bulling, A., Blanke, U. and Schiele, B., 2014. A tutorial

on human activity recognition using body-worn

inertial sensors. Computing Surveys, vol. 46, pp. 33:1-

33.

Buracchio, T., Dodge, H. H., Howieson, D., Wasserman,

D. and Kaye, J., 2010. The trajectory of gait speed

preceding mild cognitive impairment. Archives of

neurology, vol. 67, pp. 980-986.

Cola, G., Vecchio, A. and Avvenuti, M., 2014. Improving

the performance of fall detection systems through

walk recognition. Journal of Ambient Intelligence and

Humanized Computing, vol. 5, pp. 843-855.

Cuddihy, P. E., Yardibi, T., Legenzoff, Z. J., Liu, L.,

Phillips, C. E., Abbott, C., Galambos, C., Keller, J.,

Popescu, M. and Back, J., 2012. Radar walking speed

measurements of seniors in their apartments:

Technology for fall prevention. In Proc. 34th Annual

International Conference of the IEEE Engineering in

Medicine and Biology Society (San Diego, CA, USA),

pp. 260-263.

Egerton, T., Thingstad, P. and Helbostad, J. L., 2014.

Comparison of programs for determining temporal-

spatial gait variables from instrumented walkway data:

PKmas versus GAITRite. BMC research notes, vol. 7

pp. 1-7.

Fritz, S. and Lusardi, M., 2009. White Paper: "Walking

Speed: the Sixth Vital Sign. Journal of Geriatric

Physical Therapy, vol. 32, pp. 2–5.

Hamm, J., Money, A. G., Atwal, A. and Paraskevopoulos,

I., 2016. Fall prevention intervention technologies: A

conceptual framework and survey of the state of the

art. Journal of Biomedical Informatics, vol. 59, pp.

319–345.

Jian, Q., Yang, J., Yu, Y., Björkholm, P. and Mckelvey,

T., 2014. Detection of Breathing and Heartbeat by

Using a Simple UWB Radar System. In Proc. 8th

European Conference on Antennas and Propagation

(The Hague, The Netherlands), pp. 3078-3081.

Liu, L., Popescu, M., Ho, K. C., Skubic, M. and Rantz,

M., 2012. Doppler Radar Sensor Positioning in a Fall

Detection System. 34th Annual International

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors

137

Conference of the IEEE EMBS (San Diego, California

USA, 28 Aug. – 1 Sep., 2012).

Luque, R., Casilari, E., Morón, M.-J. and Redondo, G.,

2014. Comparison and Characterization of Android-

Based Fall Detection Systems. Sensors, vol. 14, pp.

18543-18574.

Lusardi, M., 2012. Is Walking Speed a Vital Sign? Topics

in Geriatric Rehabilitation, vol. 28, pp. 67–76.

Miękina, A., Wagner, J., Mazurek, P. and Morawski, R.

Z., 2016a. Selected algorithms for measurement data

processing in impulse-radar-based system for

monitoring of human movements. In Proc. IMEKO

TC1-TC7-TC13 Joint Symposium (Berkeley, CA,

USA), pp. 1-6.

Miękina, A., Wagner, J., Mazurek, P., Morawski, R. Z.,

Sudmann, T. T., Børsheim, I. T., Øvsthus, K.,

Jacobsen, F. F., Ciamulski, T. and Winiecki, W.,

2016b. Development of software application dedicated

to impulse-radar-based system for monitoring of

human movements. In Proc. IMEKO TC1-TC7-TC13

Joint Symposium (Berkeley, CA, USA), pp. 1-6.

Morawski, R. Z., Yashchyshyn, Y., Brzyski, R., Jacobsen,

F. and Winiecki, W., 2014. On applicability of

impulse-radar sensors for monitoring of human

movements. In Proc. IMEKO TC-4 International

Symposium (Benevento, Italy), pp. 786–791.

Stone, E., Skubic, M., Rantz, M., Abbott, C. and Miller,

S., 2015. Average in-home gait speed: Investigation of

a new metric for mobility and fall risk assessment of

elders. Gait & posture, vol. 41, pp. 57-62.

Studenski, S., Perera, S., Patel, K., Rosano, C., Faulkner,

K., Inzitari, M., Brach, J., Chandler, J., Cawthon, P.,

Barrett Connor, E., Nevitt, M., Visser, M.,

Kritchevsky, S., Badinelli, S., Harris, T., Newman, A.

B., Cauley, J., Ferrucci, L. and Guralnik, J., 2011. Gait

Speed and Survival in Older Adults. Journal of the

American Medical Association vol. 305, pp. 50-58.

Su, B. Y., Ho, K. C., Rantz, M. and Skubic, M., 2015.

Doppler Radar Fall Activity Detection Using The

Wavelet Transform. IEEE Transactions on Biomedical

Engineering, vol. 62, pp. 865-875.

Thingstad, P., Egerton, T., Ihlen, E. F., Taraldsen, K.,

Moe-Nilssen, R. and Helbostad, J. L., 2015.

Identification of gait domains and key gait variables

following hip fracture. BMC geriatrics, vol. 15, pp. 1-

Thong, Y. K., Woolfson, M. S., Crowe, J. A., Hayes-Gill,

B. R. and Jones, D. A., 2004. Numerical double

integration of acceleration measurements in noise.

Measurement, vol. 36, pp. 73–92.

Tomii, S. and Ohtsuki, T., 2012. Falling Detection Using

Multiple Doppler Sensors. In Proc. 14th IEEE

International Conference on e-Health Networking,

Applications and Services (Beijing, China), pp. 196–

201.

Wagner, J., Mazurek, P. and Morawski, R. Z., 2015.

Regularised Differentiation of Measurement Data. In

Proc. XXI IMEKO World Congress "Measurement in

Research and Industry" (Prague, Czech Republic), pp.

1-6.

HEALTHINF 2017 - 10th International Conference on Health Informatics

138