Feature Selection Evaluation for Light Human Motion Identification

in Frailty Monitoring System

Evangelia Pippa

, Iosif Mporas

1,2

and Vasileios Megalooikonomou

Multidimentional Data Analysis and Knowledge Discovery Laboratory,

Dept. of Computer Engineering and Informatics, University of Patras, Rion-Patras, Greece

School of Engineering and Technology, University of Hertfordshire, Hatfield, U.K.

Keywords: Human Motion Detection, Machine Learning, Feature Extraction.

Abstract: In order to plan and deliver health care in a world with increasing number of older people, human motion

monitoring is a must in their surveillance, since the related information is crucial for understanding their

physical status. In this article, we focus on the physiological function and motor performance thus we

present a light human motion identification scheme together with preliminary evaluation results, which will

be further exploited within the FrailSafe Project. For this purpose, a large number of time and frequency

domain features extracted from the sensor signals (accelerometer and gyroscope) and concatenated to a

single feature vector are evaluated in a subject dependent cross-validation setting using SVMs. The mean

classification accuracy reaches 96%. In a further step, feature ranking and selection is preformed prior to

subject independent classification using the ReliefF ranking algorithm. The classification model using

feature subsets of different size is evaluated in order to reveal the best dimensionality of the feature vector.

The achieved accuracy is 97% which is a slight improvement compared to previous approaches evaluated

on the same dataset. However, such an improvement can be considered significant given the fact that it is

achieved with lighter processing using a smaller number of features.

1 INTRODUCTION

The ageing population around the world is

increasing and it is estimated that two billion people

will be aged over 65 years by 2050. This will affect

the planning and delivery of health and social care as

well as the clinical condition of frailty. Frailty is a

medical syndrome which is characterized by

diminished strength, endurance, and reduced

physiologic function that increases an individual’s

vulnerability for developing increased dependency

and/or death (Morley et al., 2013). Frailty is

characterized by multiple pathologies such as weight

loss, weakness, low activity, slow motor

performance, balance and gait abnormalities, as well

as cognitive ones (Chen et al., 2014). Frailty

increases risks of incident falls, worsening of

mobility, disability, hospitalization or

institutionalization, and mortality (Abellan et al. ,

2008; Mitnitski et al., 2002; Morley et al., 2006),

which in turn increase the burden to cares and costs

to the society.

It is assumed that early intervention with frail

persons will improve quality of life and reduce

health services costs. Thus it is essential to develop

real life tools for the assessment of physiologic

reserve and the need to test interventions that alter

the natural course of frailty since frailty is a dynamic

and not an irreversible process. Several efforts have

been done in this direction through research and

development activities. In (Seacw Project) an

Ecosystem for training, informing and providing

tools, processes, methodologies for ICT and active,

healthy aging was developed mainly targeting to

caregivers, older people and general population. In

(Eldergames Project) an interactive tabletop

platform able to integrate potentialities derived from

both technology and leisure activities was designed.

Another purpose of (Eldergames Project) was the

monitoring of the older people status, with

information about his/her progression /regression in

cognitive health. In (Kinoptim Project) a home-care

solution that will address older people living in the

community in a preventive manner and rely on ICT

and virtual reality gaming through the exploitation

of haptic technologies, vision control and context

Pippa, E., Mporas, I. and Megalooikonomou, V.

Feature Selection Evaluation for Light Human Motion Identiﬁcation in Frailty Monitoring System.

In Proceedings of the International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2016), pages 88-95

ISBN: 978-989-758-180-9

Figure 1: The FrailSafe conceptual diagram for motion monitoring.

awareness methods was developed and integrated,

while promising to redefine fall prevention by

motivating people to be more active, in a friendly

way and with tele-supervision if necessary. In

(Mporas et al., 2015) a more holistic, personalized,

medically efficient and economical monitoring

system for people with epilepsy was provided. In

(Doremi Project) multidisciplinary research areas in

serious games, social networking, Wireless Sensor

Network, activity recognition and contextualization,

behavioral pattern analysis were combined in pilot

setups involving both older users and care providers.

In (Alfred Project) a mobile, personalized assistant

for older people was developed, using cutting edge

technologies such as advanced speech interaction,

which helps them stay independent, coordinate with

carers and foster their social contacts. In (Home

Sweet Home Project) new, economically sustainable

home assistance service which extends older people

independent living was introduced, measuring the

impact of monitoring, cognitive training and e-

Inclusion services on the quality of life of older

people, on the cost of social and healthcare delivered

to them, and on a number of social indicators. In

(Mobiserv Project) the objective was to develop and

test a proactive personal robotic, integrated with

innovative sensors, localization and communication

technologies, and smart textiles to support

independent living for older adults, in their home or

in various degrees of institutionalization, with a

focus on health, nutrition, well-being, and safety. In

(Fate Project) innovative ICT-based solution for the

detection of falls in ageing people were studied,

covering prevention and detection of falls in all

circumstances.

Human motion monitoring is a must in

surveillance of older people, since the related

information is crucial for understanding the physical

status and the behavior of the older people. This is

typically achieved using image/video and

accelerometer based data (Foroughi et al., 2008;

Xiang et al., 2015;Yang et al., 2010). In this article

we focus on the physiological function and motor

performance thus we present a light human motion

identification scheme together with preliminary

evaluation results, which will be further exploited

within the FrailSafe (Frailsafe Project) architecture.

The main contributions of this paper are summarized

in the following:

 the classification accuracy in the examined

dataset is slightly improved (about one unit)

compared to previous approaches.

 a lighter human motion identification module

using less features but achieving equal

accuracy in comparison to the one previously

proposed for the dataset under consideration is

achieved by feature selection

The reminder of this article is organized as follows.

In Section 2 we present the FrailSafe concept. In

Section 3 we describe the human motion

identification scheme. In Sections 4 and 5 we

present the experimental setup and the evaluation

results respectively. Finally, in Section 6 we

conclude this work.

2 THE FRAILSAFE CONCEPT

FrailSafe aims to better understand frailty and its

relation to co-morbidities, to develop quantitative

and qualitative measures to define frailty and to use

these measures to predict short and long-term

outcome. In order to achieve these goals real life

tools for the assessment of physiological reserve and

Feature Selection Evaluation for Light Human Motion Identiﬁcation in Frailty Monitoring System

Figure 2: Human Motion Identification Module.

of external challenges will be developed. These tools

will provide an adaptive model (sensitive to

changes) in order that pharmaceutical and non-

pharmaceutical interventions, which will be

designed to delay, arrest or even reverse the

transition to frailty. Moreover, FrailSafe targets at

creating "prevent-frailty" evidence based

recommendations for older people regarding

activities of daily living, lifestyle and nutrition, as

well as strengthening the motor cognitive, and

other"anti-frailty" activities through the delivery of

personalized treatment programs, monitoring alerts,

guidance and education. The FrailSafe conceptual

infrastructure for motion monitoring is illustrated in

Figure 1.

Through patient-specific interventions, FrailSafe

aims to define a frailty measure. This measure is

initially constructed from prior knowledge on the

field, and then globally updated based on analysis of

long-term observations of all older peoples' states.

This update is then applied to the individual patient

models, modifying them accordingly, to fit different

needs per patient. The monitoring of the older

people's motion activity is performed through the

environmental sensors module, which includes

accelerometer sensors for the monitoring of the

human motions. Details about the light motion

identification implementation are provided in the

next section.

3 LIGHT HUMAN MOTION

IDENTIFICATION

The presented workflow for light human motion

identification is part of an end-to-end system for

sensing and predicting treatment of frailty and

associated co-morbidities using advanced

personalized models and advanced interventions, the

FRAILSAFE framework.

The proposed classification methodology can be

used as a core module in order to discriminate the

detected motions to six basic activities: walking,

walking-upstairs, walking-downstairs, sitting,

standing and laying. The block diagram of the

overall workflow for learning the activity classifiers

is illustrated in Figure 2.

The multi-parametric sensor (accelerometer and

gyroscope) data are pre-processed as in (Anguita et

al., 2012; Reyes-Ortiz et al., 2013) by applying noise

filters and then sampled in fixed-width sliding

windows 



,1 (frames) of 2.56 sec and

50% overlap. The sensor acceleration signal, which

has gravitational and body motion components, was

separated using a Butterworth low-pass filter into

body acceleration and gravity. The gravitational

force is assumed to have only low frequency

components, therefore a filter with 0.3 Hz cut-off

frequency was used. From each frame, a vector of

features 



∈



,





|



| was obtained by

calculating variables from the time 





∈

|



and

frequency domain 





∈

|



The extracted time domain and frequency

domain features are concatenated to a single feature

vector as a representative signature for each frame.

Details on the type of extracted features are provided

in Section 4.

All frames are used as input to

FRAILSAFE's human motion identification module

which classifies basic activities of daily living

(ADLs) in order to obtain some preliminary

evaluation results for the proposed architecture. In

this module, a model for multiclass classification

between six basic ADLs (walking, walking-upstairs,

walking-downstairs, sitting, standing and laying),

which has been previously built in a training phase,

is used in order to label the frames. Each frame is

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

classified independently.

During the training phase of the classification

architecture, frames with known class labels (labeled

manually) are used to train the multiclass

classification model. During the test phase the

unknown multi-parametric sensor signals are pre-

processed and parameterized with similar setup as in

the training phase. Each extracted feature vector is

provided as input to the trained classifier.

4 EXPERIMENTAL SETUP

4.1 Data

The previously described classification methodology

was evaluated on multi-parametric data from the

UCI HAR Dataset (Anguita et al., 2013). The dataset

consists of accelerometer and gyroscope recordings

from 30 volunteers within an age bracket of 19-48

years when performing six activities (walking,

walking-upstairs, walking-downstairs sitting,

standing, laying). For the experiments each person

worn a smartphone (Samsung Galaxy S II) on the

waist. Using its embedded accelerometer and

gyroscope, 3-axial linear acceleration and 3-axial

angular velocity at a constant rate of 50Hz were

captured. The data were labelled manually using the

corresponding video recordings which were captured

during the experiments. Since the evaluation here

was held using a subject dependent cross-validation

setting, data were initially merged in a single dataset

and then split in 30 datasets, one for each subject.

4.2 Feature Extraction and

Classification Algorithm

Initially, the sensor signals (accelerometer and

gyroscope) were pre-processed as proposed in

(Anguita et al., 2012, Reyes-Ortiz et al., 2013) in

order to proceed with feature extraction. The

features selected for this analysis are those proposed

in (Anguita et al., 2012, Reyes-Ortiz et al., 2013)

which come from the accelerometer and gyroscope

3-axial raw signals denoted as tAcc-XYZ and tGyro-

XYZ with prefix 't' used to denote time. The

sampling frequency of these time domain signals

was 50 Hz. In order to remove noise Anguita et al.

performed low pass filtering using a median filter

and a 3rd order low pass Butterworth filter with a

corner frequency of 20 Hz. Then, in order to

separate the acceleration signal into body and

gravity acceleration signals denoted as tBodyAcc-

XYZ and tGravityAcc-XYZ, they used another low

pass Butterworth filter with a corner frequency of

0.3 Hz.

Subsequently, Jerk signals denoted as

tBodyAccJerk-XYZ and tBodyGyroJerk-XYZ were

obtained by the time derivation of the body linear

acceleration and angular velocity. Also, they used

the Euclidean norm to calculate the magnitude of

these three-dimensional signals yielding the

following signals: tBodyAccMag, tGravityAccMag,

tBodyAccJerkMag, tBodyGyroMag and

tBodyGyroJerkMag.

Finally a Fast Fourier Transform (FFT) was

applied to signals tBodyAcc-XYZ, tBodyAccJerk-

XYZ, tBodyGyro-XYZ, tBodyAccJerkMag,

tBodyGyroMag, tBodyGyroJerkMag producing

fBodyAcc-XYZ, fBodyAccJerk-XYZ, fBodyGyro-

XYZ, fBodyAccJerkMag, fBodyGyroMag,

fBodyGyroJerkMag. Here, the prefix 'f' were used to

indicate frequency domain signals.

These signals were used to estimate variables of

the feature vector for each pattern: '-XYZ' is used to

denote 3-axial signals in the X, Y and Z directions.

The aforementioned signals which were produced by

processing accordingly the initial sensor recordings

are tabulated in Table 1.

The set of features that were extracted from these

signals are those proposed by Anguita et al. (Anguita

et al., 2012) including the mean value, the standard

deviation, the median absolute deviation, the largest

value in array, the smallest value in array, the signal

magnitude area, the energy measure as the sum of

the squares divided by the number of values, the

interquartile range, the signal entropy, the

autoregression coefficients with Burg order equal to

4, the correlation coefficient between two signals,

the index of the frequency component with largest

Table 1: Pre-processed Signals.

Signals

tBodyAcc-XYZ

tGravityAcc-XYZ

tBodyAccJerk-XYZ

tBodyGyro-XYZ

tBodyGyroJerk-XYZ

tBodyAccMag

tGravityAccMag

tBodyAccJerkMag

tBodyGyroMag

tBodyGyroJerkMag

fBodyAcc-XYZ

fBodyAccJerk-XYZ

fBodyGyro-XYZ

fBodyAccMag

fBodyAccJerkMag

fBodyGyroMag

fBodyGyroJerkMag

Feature Selection Evaluation for Light Human Motion Identiﬁcation in Frailty Monitoring System

Table 2: Additional Signals.

Additional Siganls

gravityMean

tBodyAccMean

tBodyAccJerkMean

tBodyGyroMean

tBodyGyroJerkMean

magnitude, the weighted average of the frequency

components to obtain a mean frequency, the

skewness of the frequency domain signal, the

kurtosis of the frequency domain signal, the energy

of a frequency interval within the 64 bins of the FFT

of each window and the angle between two vectors.

Additional vectors were obtained by averaging

the signals in a signal window sample. These are

used on the angle variable (Table 2).

In conclusion, for each record a 561- feature

vector with the aforementioned time and frequency

domain variables was provided.

The computed feature vectors were used to train

a classification model. In order to evaluate the

ability of the above features to discriminate between

ADLs we examined the SMO (Keerthi et al., 2001;

Platt et al., 1998) with RBF kernel classification

algorithm, which was implemented by the WEKA

machine learning toolkit (Hall et al. 2009). SMO

algorithm is an implementation of Support Vector

Machines provided by the WEKA toolkit. Here we

selected SMO for the classification since SVMs are

used mostly in relevant literature.

During the test phase, the sensor signals were

pre-processed and parameterized as during training.

The SMO classification model was used to label

each of the activities. Evaluation was performed in a

subject dependent cross-validation setting.

In a further step we examined the discriminative

ability of the extracted features for the human

motion identification. The ReliefF algorithm

(Kononenko, 1994) (which is an extension of an

earlier algorithm called Relief (Kira and Rendell,

1992)) was used for estimating the importance of

each feature in multiclass classification. In the

ReliefF algorithm the weight of any given feature

decreases if the squared Euclidean distance of that

feature to nearby instances of the same class is more

than the distance to nearby instances of the other

class. ReliefF is considered one of the most

successful feature ranking algorithms due to its

simplicity and effectiveness (Dietterich, 1997; Sun

and Li, 2006; Sun and Wu, 2008) (only linear time

in the number of given features and training samples

is required), noise tolerance and robustness in

detecting relevant features effectively, even when

these features are highly dependent on other features

(Dietterich, 1997; Kononenko,1997).

Table 3: Subject Dependent Human Motion Identification

Accuracy.

Subject Accuracy

1 100%

2 93,71%

3 97,36%

4 93,69%

5 86,09%

6 92,92%

7 93,83%

8 98,93%

9 85,76%

10 95,92%

11 98,42%

12 97,50%

13 95,41%

14 95,67%

15 96,65%

16 94,81%

17 90,76%

18 96,70%

17 98,33%

20 96,33%

21 98,53%

22 99,38%

23 96,77%

24 98,95%

25 94,87%

26 96,94%

27 99,73%

28 93,72%

29 100,00%

30 97,65%

Table 4: Mean across subjects confusion matrix. Rows represent the actual class and columns the predicted class.

Standing Sitting Laying Walking Downstairs Upstairs

Standing 1795 110 0 0 0 1

Sitting 289 1485 2 0 0 1

Laying 0 0 1944 0 0 0

Walking 0 0 0 1718 2 2

Downstairs 0 0 0 5 1397 4

Upstairs 0 0 0 0 0 1544

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

Table 5: ReliefF Feature Ranking.

Ranking Feauture

1 tGravityAcc_energy_X

2 fBodyAccJerk_entropy_X

3 fBodyAcc_entropy_X

4 fBodyAccJerk_entropy_Y

5 tBodyAccJerkMag_entropy

6 angle(X_gravityMean)

7 tGravityAcc_min_X

8 tGravityAcc_mean_X

9 tBodyAccJerk_entropy_X

10 tGravityAcc_max_X

11 fBodyBodyAccJerkMag_entropy

12 tBodyAcc_max_X

13 tBodyAccJerk_entropy_Y

14 fBodyAccMag_entropy

15 fBodyAcc_entropy_Y

16 fBodyAccJerk_entropy_Z

17 tBodyAccJerk_entropy_Z

18 tBodyGyroJerkMag_entropy

17 tGravityAcc_energy_Y

20 tBodyAccMag_entropy

21 tGravityAccMag_entropy

22 tGravityAcc_mean_Y

23 tBodyGyroJerk_entropy_Z

24 tGravityAcc_max_Y

25 fBodyAcc_entropy_Z

26 tGravityAcc_entropy_Z

27 tGravityAcc_min_Y

28 fBodyGyro_entropy_X

29 fBodyGyro_entropy_X

30 tBodyGyroJerk_entropy_X

31 fBodyAcc_mad_X

32 fBodyAcc_std_X

33 tBodyAcc_std_Y

34 fBodyAcc_mad_Y

35 tBodyAcc_std_X

36 fBodyBodyGyroJerkMag_entropy

37 tBodyAcc_mad_Y

38 fBodyGyro_entropy_Y

39 tBodyGyroMag_entropy

40 fBodyAcc_std_Y

Furthermore, ReliefF avoids any exhaustive or

heuristic search compared with conventional

wrapper methods and usually performs better

compared to filter methods due to the performance

feedback of a nonlinearclassifier when searching for

useful features (Sun and Wu, 2008).

In this study, ranking is performed on the whole

dataset including all frames from all subjects. We

examined the performance of the method, in terms

of accuracy for different number of N-best features

(N =10, 20, 30, ... 560 ), with respect to the ReliefF

feature ranking algorithm.

5 EXPERIMENTAL RESULTS

The classification methodology presented in Section

3 was evaluated using the classification algorithm

and the cross-validation scheme described in Section

4. The classification performance was evaluated in

terms of accuracy



 





(1)

where TP denotes the true positives, TN the true

negatives, FP the false positives and FN the false

negatives. The results of the method using all

features are shown on Table 3.

As can be seen in Table 3, the overall highest

accuracy of the proposed methodology for human

motion identification is 100% for subjects 1 and 29

and the lowest accuracy 85.76% was obtained for

the 9th subject. However, the mean accuracy is

relatively high 95,84%. The mean across all

subjects confusion matrix is shown in Table 4. As

can be seen all ADLs except sitting and standing are

nearly perfectly discriminated from the others with

only a few false dismissals or false alarms (2 to 5).

The misclassification of some sitting and standing

instances are probably owed to the similarity of

these ADLs.

In a further step, we applied feature ranking on

the whole dataset (consisting of all available

subjects) using the ReliefF algorithm as described in

Section 4. The performance of the classification, in

terms of accuracy, for different number of N-best

features (N =10, 20, 30, ..., 560) for the SMO

algorithm is shown in Figure 3.

As can be seen in Figure 3 the highest

classification accuracy is achieved when a large

subset of discriminative features are used.

Specifically, the highest accuracy is achieved for a

subset of 550 best features with a percentage of 97%

which is equal to the accuracy achieved when all

features are used. It seems that the size and the

variability of the dataset is relatively large requiring

a feature vector of high dimensionality to accurately

discriminate between the six classes. However, with

only 40 features a high accuracy equal to 90% can

be achieved.

Table 5 shows the 40 best features according to

the ReliefF ranking algorithm. Although it is best to

use a high dimensional feature vector to achieve

higher classification accuracy, feature selection is

still important in cases where a light human motion

identification module is needed such as in

FRAILSAFE.

Although direct comparison with other studies

Feature Selection Evaluation for Light Human Motion Identiﬁcation in Frailty Monitoring System

Figure 3: Classification Accuracy for different subsets of N-best features (N=10,20,..., 550).

performed on the same dataset is not feasible due to

different problem identification (subject dependent

classification studied here versus subject

independent classification studied in previous

works) and different validation protocols followed

as well (cross validation used here instead of 70%

and 30% train and test sets respectively in the

literature), the proposed method for the subject

independent classification slightly improves the

classification accuracy from 94% and 96% achieved

in (Reiss et al., 2013) and (Romera-Paredes et al.,

2013, Kastner et al., 2013) respectively to 97%. This

improvement is significant since it is being achieved

with less features providing the means for lighter

approaches for human motion identification.

6 CONCLUSIONS

In this paper, we investigated the problem of human

motion identification from multi-parametric sensor

data acquired from accelerometers and gyroscopes

using a large number of time-domain and frequency

domain features in order to be used as part of an

end-to-end system for sensing and predicting

treatment of frailty and associated co-morbidities

using advanced personalized models and advanced

interventions. The proposed methodology was

evaluated in multi-parametric data from 30 subjects.

The evaluation of the multiclass SMO classification

algorithm showed that a mean accuracy of 96% was

achieved. Feature ranking investigation and

evaluation of the classification models using subsets

of features were performed and revealed the most

significant features for the classification task. The

use of the most discriminative features (N = 550)

achieved accuracy equal to the accuracy achieved

when all features are used.

ACKNOWLEDGEMENTS

The research reported in the present paper was

partially supported by the FrailSafe Project (H2020-

PHC-21-2015 - 690140) “Sensing and predictive

treatment of frailty and associated co-morbidities

using advanced personalized models and advanced

interventions”, co-funded by the European

Commission under the Horizon 2020 research and

innovation programme.

REFERENCES

Abellan van Kan, G., et al., 2008. The I.A.N.A Task Force

on frailty assessment of older people in clinical

practice. J Nutr Health Aging, 12(1): p. 29-37.

Alfred Project:

http://cordis.europa.eu/project/rcn/110629_en.html.

Anguita, D., Ghio, A., Oneto, L., Parra, X., Reyes-Ortiz,

J.L., 2013. A Public Domain Dataset for Human

Activity Recognition Using Smartphones. ESANN

2013, 21th European Symposium on Artificial Neural

Networks, Computational Intelligence and Machine

Learning.

Anguita, D., Ghio, A., Oneto, L., Parra, X., Reyes-Ortiz,

J.L., 2013. Human Activity Recognition on

Smartphones using a Multiclass Hardware-Friendly

Support Vector Machine. IWAAL 2012 International

Workshop of Ambient Assisted Living.

Chen, X., Mao, G., Leng, S.X., 2014. Frailty syndrome: an

overview. Clin Interv Aging, 9: p. 433-41.

Dietterich, T. G., 1997. Machine learning research: Four

current directions. Artificial Intelligence Magazines,

vol. 18, pp. 97–136.

Doremi Project: http://cordis.europa.eu/project/rcn/110829

_en.html.

Eldergames Project: http://cordis.europa.eu/project/rcn/

80186_en.html.

Fate Project: http://cordis.europa.eu/project/rcn/191694

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

_en.html.

Foroughi H., Aski B.S., Pourreza H., 2008. Intelligent

video surveillance for monitoring fall detection of

elderly in home environments, 11th International

Conference on Computer and Information

Technology. ICCIT 2008.

Hall, M., Frank, E., Holmes, G., Pfahringer, B.,

Reutemann P., Witten, I. H., 2009. The WEKA Data

Mining Software: An Update. SIGKDD Explorations,

vol. 11.

Home Sweet Home Project: http://cordis.europa.eu/

project/rcn/191712_en.html.

Kastner, M., Strickert, M., Villmann, T., 2013. A sparse

kernelized matrix learning vector quantization model

for human activity recognition. European Symposium

on Artificial Neural Networks, Computational

Intelligence and Machine Learning (ESANN).

Keerthi, S.S., Shevade, S.K., Bhattacharyya, C., Murthy,

K.R.K., 2001. Improvements to Platt's SMO algorithm

for SVM classifier design. Neural Computation, vol.

13, pp. 637-649.

Kinoptim Project: http://cordis.europa.eu/project/rcn/

106678_en.html.

Kira, K., Rendell, L. A., 1992. A practical approach to

feature selection. Proc. 9th Int. Conf. Mach. Learn.,

pp. 249 – 256.

Kononenko, I., 1994. Estimating attributes: Analysis and

extension of RELIEF. Proc. Euro. Conf. Mach.

Learn., vol. 784, pp. 171– 182.

Kononenko, I., Simec, E., Robnik-Sikonja, M., 1997.

Overcoming the Myopic of Inductive Learning

Algorithms with RELIEF-F. Applied Intelligence.

Mitnitski, A.B., et al., 2002. Frailty, fitness and late-life

mortality in relation to chronological and biological

age. BMC Geriatr, 2: p. 1.

Mobiserv Project: http://cordis.europa.eu/project/rcn/

93537_en.html.

Morley, J.E., et al., 2006. Frailty. Med Clin North Am, .

90(5): p. 837-47.

Morley, J.E., et al., 2013. Frailty consensus: a call to

action. J Am Med Dir Assoc, 14(6): p. 392-7.

Mporas, I., Tsirka, V., Zacharaki, E.I., Koutroumanidis,

M., Richardson, M., Megalooikonomou, V., 2015.

Seizure detection using EEG and ECG signals for

computer-based monitoring, analysis and management

of epileptic patients. Expert Systems with Applications,

Volume 42, Issue 6, 15, Pages 3227–3233.

Platt, J., 1998. Fast Training of Support Vector Machines

using Sequential Minimal Optimization. Advances in

Kernel Methods - Support Vector Learning.

Reiss, A., Hendeby, G., Stricker, D., 2013. A competitive

approach for human activity recognition on

smartphones. European Symposium on Artificial

Neural Networks, Computational Intelligence and

Machine Learning (ESANN).

Reyes-Ortiz, J.L., Ghio, A., Parra, X., Anguita, D.,

Cabestany, J., Catala, A., 2013. Human Activity and

Motion Disorder Recognition: Towards Smarter

Interactive Cognitive Environments. ESANN 2013

21th European Symposium on Artificial Neural

Networks, Computational Intelligence and Machine

Learning.

Romera-Paredes, B., Aung, H., Bianchi-Berthouze, N.,

2013. A one-vs-one classifier ensemble with majority

voting for activity recognition. European Symposium

on Artificial Neural Networks, Computational

Intelligence and Machine Learning (ESANN).

Seacw Project: http://cordis.europa.eu/project/rcn/191786

_en.html.

Sun, Y., Li, J. 2006. Iterative RELIEF for feature

weighting. Proc. 21st Int. Conf. Mach. Learn., pp.

913–920.

Sun, Y., Wu, D., 2008. A RELIEF based feature extraction

algorithm. In Proceedings of SIAM International

Conference on Data Mining.

Xiang, Y., Tang, Y., Ma, B., Yan, H., Jiang, J., Tian, X.,

2015. Remote Safety Monitoring for Elderly Persons

Based on Omni-Vision Analysis, PLoS One, 10(5).

Feature Selection Evaluation for Light Human Motion Identiﬁcation in Frailty Monitoring System