Rehabilitation for Children while Playing with a Robotic Assistant in a

Serious Game

L.V. Calderita

, P. Bustos

, C. Su

arez Mej

ıas

, B. Ferrer Gonz

alez

and A. Bandera

RoboLab, University of Extremadura, C

aceres, Spain

GIT, UCAi, Hospital Universitario Virgen del Roc

ıo, Sevilla, Spain

Grupo ISIS, University of M

alaga, M

alaga, Spain

Keywords:

Neuro-rehabilitation Therapies, Socially Assistant Robotics, Augmented Reality, Serious Games, Paediatric

Patients.

Abstract:

Traditional neuro-rehabilitation therapies are usually repetitive and lengthy, reducing motivation and adher-

ence to the treatment and thus limiting the beneﬁts for the patients. Moreover, exercises are usually not

customizable for the patients, further increasing their disengagement with the treatment. The outcome is then

a boring session day after day. This is more pronounced when the patient is a child. However, the execution of

these repetitive movements is really needed, as it alters the properties of our neurons, including their pattern

of connectivity. Correctly driven, this process ﬁnally allows to improve the neural functionality. The question

is then: how can we improve the motivation and immersion of the patients into the therapy? We could try to

convert the boring therapy into a funny one. This will help to the patients, but also to the practitioner. For

this end, computer-assisted technologies have been extensively employed in the last years. Within this re-

search ﬁeld, this paper proposes to engage the child to the therapy by immersing her into an augmented reality

scenario, where it will play several serious games. The adherence to the session will be further increased by

incorporating a social robot as a playmate. This robot will be a personal trainer, that will perform the session

in the real world with the patient. Additionally, the robot will be able to record the data for each session. This

data could be subsequently used by the rehabilitation specialists for monitoring and/or adapting the therapy to

the patient’s needs.

1 INTRODUCTION

Exploiting the use-dependent plasticity of our neu-

romuscular system, neuro-rehabilitation therapies are

devised to help patients with some motor impairment.

These therapies take advantage of the fact that the

motor activity alters the properties of our neurons,

including the pattern of their connectivity, and thus

their functionality (Leocani and Comi, 2006). Hence,

a sensor-motor treatment where the patient makes cer-

tain movements, will help him to (re)learn how to

move the affected body parts. Because lack of arm-

movement control directly affects activities of daily

living and independence (Whitall et al., 2000), this

improving of the upper-limb motor function is of

great importance.

The basis of the rehabilitation process is the repe-

tition of certain movements, being the recovery corre-

lated with the frequency and intensity of these move-

ments. On the contrary, passive movements -postures-

are insufﬁcient to alter motor recovery. Hence, the fo-

cus of the rehabilitation should be on movement co-

ordination (active) rather than muscle strengthening

(passive) (Hermano and Hogan, 2009). This tradi-

tional rehabilitation process comes at a cost: therapies

are usually repetitive and lengthy, reducing motiva-

tion and adherence to the treatment and thus limiting

the beneﬁts for the patients (Steultjens et al., 2003).

Clinical experiments demonstrate that motivation

is an important factor for successfully addressing a

lengthy neuro-rehabilitation therapy and it is usu-

ally employed as a determinant of rehabilitation out-

come (Colombo et al., 2007). Hence, active engage-

ment towards a therapy is typically equated with mo-

tivation. Technology-assisted training can provide

engaging and task-oriented training using patient-

tailored feedback to support the (re)learning of motor

skills (Timmermans et al., 2009). From pioneering

systems such as the Lokomat from Hocoma (Jezernik

et al., 2003), the application of computer-assisted

V. Calderita L., Bustos P., Suárez Mejías C., Ferrer González B. and Bandera A..

Rehabilitation for Children while Playing with a Robotic Assistant in a Serious Game.

DOI: 10.5220/0004646700890096

In Proceedings of the International Congress on Neurotechnology, Electronics and Informatics (VirtRehab-2013), pages 89-96

ISBN: 978-989-8565-80-8

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

technologies to rehabilitation has generated a pos-

itive feedback from therapists. For instance, the

ArmeoSpring, a more recent proposal from Hocoma,

is a robotic tool to improve therapy by facilitating in-

tensive and functional movement exercises. As it is

proposed by Colombo et al, this tool supports the ther-

apy by motivating, game-like tasks (Colombo et al.,

2007). Video games have long been known to be en-

gaging to play. Thus, if rehabilitation games with

a similar degree of engagement are created, it will

be possible to improve the therapeutic results. For

this end, gaming consoles that combine entertainment

and exercise such as the Nintendo Wii or the Sony

Eyetoy can be employed. On the contrary, commer-

cial games could not be useful for people with mo-

tor function problems. They are often too fast and

frequently provide negative feedback when they are

lost (Burke et al., 2010). The design of rehabilita-

tion games requires the a priori deﬁnition of the spe-

ciﬁc proﬁle of the patient and the rehabilitation ob-

jectives. Some of these rehabilitation games employ

virtual (augmented) reality technology to immerse the

patient in a virtual scenario. For motor function reha-

bilitation, it is also common to incorporate technology

to track the movements of the patient. This tracked

data can be then used to drive a graphical represen-

tation of the patient (or a part of her) in the virtual

world. The advantages of this scenario are twofolds:

it enables the patient to achieve a high degree of con-

trol onto her activity on the game; and it improves

the degree of engagement of the game. Both issues

improve the rehabilitation therapy, increasing the pa-

tient’s control of her movements (there is a goal like

in functional-based therapies) or her motivation.

This paper proposes to combine the engagement

capabilities shown by rehabilitation games with hand-

off assistive robotics. Given the inherent people ten-

dency to engage with life-like social behaviour, the

use of the robot for augmenting or maintaining the

patient’s motivation provides an important advantage

over game-based approaches (Fasola and Mataric,

2011). Thus, socially assistive robots emerge as a

new ﬁeld of robotics whose aim is to develop systems

that assist patients through social rather than phys-

ical interaction (Tapus et al., 2007). They provide

therapy oversight, coaching and motivation using the

robot’s abilities to interact and maintain the interest

of patients. These robots are described as an intersec-

tion of assistive robotics (those that provide assistance

to a person) and socially interactive robotics (those

that communicate with people through social, non-

physical interaction) (Feil-seifer and Matari

c, 2011).

We have developed a system that uses low-cost gam-

ing and robotics technologies for the rehabilitation of

paediatric patients with upper-limb motor deﬁcit due

to cerebral palsy or brachial plexus palsy (obstetric),

but without signiﬁcant cognitive or communicative

deﬁcits.

1.1 Motivation

Cerebral palsy is a neurological chronic impairment

usually caused by a prenatal brain defect or by brain

injury during birth, that has a speciﬁc inﬂuence in

certain motor areas. It can appear in the ﬁrst gesta-

tion day or within the ﬁrst three or ﬁve years, mani-

festing with several symptoms including muscle tone,

posture and movement disorders. In addition, cogni-

tive impairments, communicative disorders, and con-

vulsive seizures (epilepsy) may be present. The inci-

dence of cerebral palsy is about 2 per 1000 live-births

in developed countries and slightly greater, about 2.5

per 1000, in developing countries. Longer prevalence

has been detected due to the increase in survival rate

in children born with low weight or other risk fac-

tors such as premature, maternal-child malnutrition

and having a pregnancy with low control. This large

incidence has an important impact on the clinical re-

sources. For instance, in 2010, 1.135 new patients

asked for a ﬁrst session on paediatric rehabilitation

due to this pathology, and a total of 2.957 patients

where attended at the Hospital Universitario Virgen

del Roc

ıo (HUVR) in Seville. On the other hand, ob-

stetrical brachial paralysis is deﬁned as a ﬂaccid pare-

sis of an upper extremity due to traumatic stretching

of the brachial plexus received at birth, with the pas-

sive range of motion greater than the active range mo-

tion (see Fig. 1)

. Brachial palsy is a paralysis in-

volving the muscles of the upper extremity that fol-

lows mechanical trauma to the spinal roots of C5 to

T1 during birth. Injuries are transient, with full re-

turn of function occuring in 70-92 % of cases (Mich-

elow et al., 1994). In Spain, we ﬁnd about 160-230

new cases of this pathology per year, which will be

associated to other visceral, vascular or cranial in-

juries. There is certain stability in the incidence of

this pathology in recent years, but this ranges from

0.5 to 1.9 per 1000 live-births.

In order to ensure that children suffering from

these two pathologies achieve the highest level of re-

covery possible, it is essential that they start sched-

uled physical therapy sessions as soon as possible.

These sessions should also be regularly conducted (in

an ideal case, it would be desirable that each patient

will be treated every day). However, both issues are

not always possible due to the lack of therapists. In

fact, at HUVR, these patients are usually treated one

http://catthsu.com/medlegal.html

NEUROTECHNIX2013-InternationalCongressonNeurotechnology,ElectronicsandInformatics

time per week. Also, it must be noted that the degree

of affectation of each patient is very different accord-

ing to both the seriousness of the disease and the bod-

ily functions affected, so it is essential that rehabilita-

tion sessions will be personalised. However, the con-

ventional rehabilitation treatment for these patholo-

gies is usually based on the repetition of a set of te-

dious exercises. This is a problem for paediatric pa-

tients due to their young age, as they would prefer do-

ing fun exercises instead of repeating the same move-

ments for twenty minutes. The resulting loss of moti-

vation can be a serious obstacle for the therapy.

Figure 1: Shoulder dystocia.

The described scenario is then suitable for the ap-

plication of new therapies based on a socially assis-

tive robot, being the use case where our robot will un-

fold its abilities. Thus, the robot acts as a coacher in

the session, explaining the exercises to the child (’we

will now play to’) and providing positive messages

through verbal and non-verbal channels. Games will

be based on an augmented reality framework, where

the entire body of the child will be projected inside a

virtual world. Within this world, the child could wear

a superhero themed uniform and will be encouraged

to perform speciﬁc exercises. In the pilot study pre-

sented in this paper, we investigated the effects of this

system on motor recovery but also on acceptance and

satisfaction grade from the patients and medical staff.

1.2 Organization of the Paper

The rest of the paper is organized as follows: Sec-

tion 2 describes the involved technologies and the pro-

posed therapy. Experimental results are presented at

Section 3. Finally, Section 4 draws the main conclu-

sions and brieﬂy introduce future work.

2 METHODOLOGY

2.1 Involved Technologies

Traditional rehabilitation therapies may be tedious.

The problem is even worse when the patient is a child.

The medical staff in this project has reported us its ex-

perience with children and adults. An immediate ob-

servation is that the effort of the practitioner to main-

tain the attention and motivation must be greater when

the patients are children. Thus, if the therapist works

with paediatric patients, she will usually need to use

little games or toys to keep their adherence to the

treatment. Within this framework, we have provided

the physiotherapists with a new tool: Ursus. Ursus

is a robot designed to conduct rehabilitation exercises

with children, developed by the RoboLab Group with

the collaboration of the Hospital Universitario Virgen

del Roc

ıo (Seville). It is a low-cost design in which a

static torso holds two robotic arms with ﬁve degrees

of freedom (DoF) each, a three DoF neck and an ar-

ticulated mouth. The head is capable of generating

simple emotions. The full platform is composed of 14

DoF and is 140 cm tall. After several revisions, it cur-

rently looks just like a big teddy bear (see Fig. 2). The

ﬁnal goal for Ursus is to make the patient move the af-

fected upper limb in each therapeutic session accord-

ing to a predeﬁned plan. These movements should

be as correct as possible and repeated with a proper

cadence.

In order to achieve this goal, Ursus will try to en-

gage the child in the game. These therapeutic games

are more than just entertainment, being their main

purpose that the patient performs speciﬁc movements

for rehabilitation (Rego et al., 2010). As described

in Section 1, games can try to improve the immer-

sion of the player into the action by using virtual or

augmented reality. To help and encourage the child

to play the game, we propose to incorporate to this

scenario a robotic playmate. Ursus will explain to

the child how to play using synthetic speech and will

project on a screen an Augmented Reality (AR) game.

In this game the child is the main actor and is encour-

aged to perform speciﬁc tasks such as grabbing some

fruits from a tree or throwing them to a basket. A

snapshot of the real scenario is illustrated in Fig. 3.

Ursus is equipped with a Red-Green-Blue and Depth

(RGBD) sensor, such as a Kinect from Microsoft or a

Xtion from Asus (Khoshelham and Elberink, 2012)

and with speakers. All software and hardware com-

ponents are running on-board within a conventional

laptop powered by the RoboComp robotics frame-

work (Manso et al., 2010) (see (Mej

ıas et al., 2013)

for further details about Ursus).

RehabilitationforChildrenwhilePlayingwithaRoboticAssistantinaSeriousGame

Figure 2: Ursus: a robotic platform for neuro-rehabilitation

therapies.

Although we could use the capacity of the OpenNI

library for tracking human bodies, we have integrated

this ability within a model-based approach for human

motion capture. We provide some results on Section

3 (see (Calderita et al., 2013) for further details). The

aim is to employ the chest or the torso of the child

like a visual landmark of augmented reality. Then, we

build a virtual scenario around the child for each re-

habilitation session (see Section 2.2 for further details

about the games). As we a priori knows the proﬁle of

the patient, each game can be easily customizable.

2.2 The proposed Therapy

As Figure 3 shows, when the exercise begins there is

one child in front of Ursus. The child motion is cap-

tured using the depth channel of the RGBD sensor

and the previously commented approach. This hu-

man motion capture algorithm is based on the OpenNI

library and provides the position and angles of the

joints of the person under analysis. This informa-

tion is further ﬁltered using a kinematic model of the

child (Calderita et al., 2011; Calderita et al., 2013) to

avoid unreal poses and changing bone lengths. The

accuracy of the system is enough to deal with the task

since the medical staff does not evaluate more than

ten degrees of freedom.

Currently, our serious games are very simple and

they only try to encourage the correct performing of

certain movements. Nevertheless, the simplicity of

these games proved to be valid for the patients in our

experimental evaluations (children whose age ranges

from 3 to 7 years old). In these games, Ursus controls

the device that projects the real image of the child in-

side of a virtual scenario. This image is also taken

from the RGBD sensor but now using the color chan-

nels (Khoshelham and Elberink, 2012). In real time,

Ursus merges both sources of information to create

on the screen a, so called, augmented scenario. It is

over this video sequence where the virtual world is

projected. The game virtually dresses the child with

a red-and-blue themed uniform, much in the spirit of

Superman or Spiderman suits. This uniform partially

covers the limbs and torso of the child. On the ﬂoor,

close to the child, there is a big tree with apples next

to a basket. Up in the air, an enormous apple is in-

ﬂated and deﬂated, seeking to capture the attention of

the patient. It should be noted that the location of all

objects inside the virtual scenario is relative to the po-

sition of the patient, as she is the reference landmark.

The child is verbally encouraged by Ursus to take

the apple and put it into the basket. To this end, Ursus

uses a probabilistic grammar to generate the adequate

sentences in real-time. This algorithm takes into con-

sideration the time from the beginning of the session,

the number of repetitive movements performed and

the current state of the interactive game. The aim is

to encourage and help the child to pass the current

level of the game. We consider that the child’s moti-

vation will be greater if she thinks that Ursus is a real

playmate. And this feeling is enforced by ’humaniz-

ing’ the human-robot interaction process. Thus, Ur-

sus synchronizes the speech generation with current

movements of its lips (Cid et al., 2011) and also with

correct non-verbal gestures (e.g. when it says ’yes’

or ’no’, it simultaneously enforces the sentence with

a current motion of its head).

The game suggests repetitive exercises to the child

and Ursus encourages her to do these movements

through verbal and non-verbal (it also performs the

movements) cues. It is a real playmate. Taken into

account the patient’s proﬁle and how she is respond-

ing to the current session, the difﬁculty of the game

can be adapted by the practitioner through a simple

control panel. This panel also allows the practitioner

to change the boundary of the game, to select other

games or to show to the patient previously recorded

videos with the exercises to perform. Finally, this

control panel also allows to visualize the patient from

the RGB camera that is mounted on Ursus. Further-

more, all data is recorded by Ursus, allowing to the

medical professionals the off-line visualization of the

session. This off-line monitoring (see Fig. 4) of the

patient’s movements is displayed using a graphic in-

terface (GUI) that not only provides the video se-

quence, but also numeric information about the am-

plitude (in milimeters) of the movements.

see a more extended example of the GUI at https://

www.youtube.com/watch?v=3NsYDbwsBYs

NEUROTECHNIX2013-InternationalCongressonNeurotechnology,ElectronicsandInformatics

Figure 3: Ursus, patient and AR-based serious game in action. The picture was taken at the Hospital Virgen del Roc

ıo (Seville,

Spain) and provides a good snapshot of the sessions with reals patients.

3 EXPERIMENTAL EVALUATION

3.1 Participants

This work focuses on the rehabilitation of paediatric

patients with upper-limb motor deﬁcit due to cerebral

palsy or brachial plexus palsy (obstetric), but with-

out signiﬁcant cognitive or communicative deﬁcits.

In order to evaluate the proposed therapy, an exper-

imental group of six paediatric patients was chosen

by the medical staff of the Department of Rehabili-

tation of the Hospital Universitario Virgen del Roc

ıo

at Seville (Spain). As aforementioned, the age of the

children ranges from three to seven years old. They

present upper-limb motor deﬁcit due to cerebral palsy

or brachial plexus palsy.

3.2 Rehabilitation Objectives and

Preliminary Results

Ursus performed several rehabilitation sessions with

this group of patients. The session was presented to

the child more like a game with a robotic friend than

a repetition of exercises. In any case, the mandatory

movements demanded by the games include shoulder

ﬂexion and abduction, elbow ﬂexion and extension,

wrist ﬂexion and extension, and forearm pronation

and supination. There was always a therapist super-

vising the session. In an on-line fashion, she was able

to perceive how the patient made the movements us-

ing the information provided by Ursus. This monitor-

ing allowed the therapist to personalise the treatment

and determine the evolution of the recovery. Showing

calmly the correct movements with his arms, talking

about interesting matters for the child, playing mu-

sic and projecting pictures, videos and augmented re-

ality (AR) games on an external screen, were some

of the resources that Ursus pull out to capture the

child’s attention and interest. It is important to note

that the main difﬁculty here was to detect the atten-

tional state of the patient when using each resource,

in order to correctly decide what to do next. Games

based on augmented reality technologies were a nat-

ural extension for Ursus, which was always tracking

the patient’s silhouette.

The clinical variables that were used for evalu-

ating the clinical evolution of the patient were pas-

sive and active articular balance of the shoulder, el-

bow and hand; degree of concordance (i.e. precision

of the movements performed by the child with re-

spect to theoretical values); motor function of upper-

limbs (’Nine Hole Peg Test’) and patients’ satisfac-

tion (’Goal Attainment Scale’). However, the val-

idation methodology of the therapy must also con-

sider, in our case, metrics related to human-robot in-

teraction. These metrics should quantify the level of

attention and engagement between robot and child.

In this work, qualitative results were obtained from

several polls of all the participants in the experiment

(paediatric patients, parents and technical and medi-

cal staff). These polls were conducted before and af-

ter the sessions and the answers were classiﬁed de-

RehabilitationforChildrenwhilePlayingwithaRoboticAssistantinaSeriousGame

Figure 4: The visualization of the session in the off-line mode.

pending on the satisfaction level of the experience.

From them, it can be concluded that the physical ap-

pearance of Ursus was quite satisfactory and that pa-

tients enjoyed the rehabilitation session and they con-

sidered it more fun and motivating than only using

the conventional treatment. Moreover, the medical

staff also considered the rehabilitation session posi-

tive for the children rehabilitation process, and the re-

sults recorded by the robot very useful for analysing

the evolution of the patients and planning personal-

ized future rehabilitation sessions. Brieﬂy, it can be

concluded that Ursus was able to achieve a high level

of engagement by the patient, maintaining the levels

of motivation and adherence to the treatment.

On the other hand, one of the main contributions

of this work is the capability of the robot Ursus to

measure and record the whole therapeutic session. As

aforementioned, we employ the method recently pro-

posed by the authors in (Calderita et al., 2013) to

capture the patient motion. Basically, we propose to

use a human model to constraint the body motion to

reachable and valid positions. It ﬁlters the deviations

of the 3D position of several relevant body parts as

model limb lengths converge toward stable values in

the learning phase (Calderita et al., 2013). Thus, even

although the OpenNI tracker provides 3D centroids

for all body parts in all processed frames, these val-

ues are modiﬁed as the model adopts, for every rele-

vant body part, the 3D position that is closest to the

centroid provided by OpenNI, but constrained by the

human model kinematics. Using this approach, the

angular positions of the patient’s limbs are recorded

in real-time with a precision of a few degrees. As Ta-

ble 1 shows for the right elbow, the joints positions are

also recorded with a precision of only few centime-

ters. Results depicted in Table 1 show that the pro-

Table 1: Mean errors and standard deviations of the right

elbow, in centimetres (Calderita et al., 2013).

Mean err Std dev

OpenNI centroids 6.7 cm 2.7 cm

Fixed limb lengths 7.2 cm 2.3 cm

Adaptive limb lengths 4.9 cm 1.7 cm

posed method improves the accuracy of the OpenNI

tracker, decreasing both elbow mean error and stan-

dard deviation.

4 CONCLUSIONS AND FUTURE

WORK

In this paper, we describe our short-time experiences

on the rehabilitation of paediatric patients with upper-

limb motor deﬁcit due to cerebral palsy or brachial

plexus palsy (obstetric). The idea of using a social

robot as a playmate increased the motivation and in-

terest of the children in the rehabilitation sessions,

showing us that the acceptance of Ursus was not a

problem in this scenario. It was very nice to see how

one of the children, after conducting the session, ap-

proached to Ursus to give it a hug. On the other

hand, the acquisition of data information about the

session was conducted on real-time, in a very pre-

cise way (see (Calderita et al., 2013) for further de-

tails about our approach for human motion capture).

Furthermore, this data can be replayed and analysed

at any time by the specialists, allowing them to com-

pare the evolution of the patient, in each session, in a

new quantitative way. We expect to improve with this

methodology and techniques the current procedures

based on standard tests.

Future work focuses on extending these experi-

NEUROTECHNIX2013-InternationalCongressonNeurotechnology,ElectronicsandInformatics

Figure 5: Ursus interacting with a child.

ences, designing new interactive AR games specif-

ically aimed at generating therapeutic movements

while providing a stimulating and joyful experience.

Finally, during the sessions, the robot will register all

movements made by the child and will compare these

movements with the normalized patterns deﬁned by

the physicians. The computed difference should be

used to generate on-line reinforcement verbally syn-

thesized discourses and expressions. It is important

to consider that, despite we have presented a rather

simple scenario, the complexity of keeping the child’s

attention during twenty minutes each session and dur-

ing tens of sessions, is daunting. Only a well designed

robot with a suitable cognitive architecture and the

knowledge of trained clinicians provides the neces-

sary material to pursue this kind of research.

ACKNOWLEDGEMENTS

This work has been partially granted by the

Spanish Ministerio de Econom

ıa y Competitividad

(MINECO) and FEDER funds under coordinated

project no. TIN2012-38079.

REFERENCES

Burke, J., McNeill, M., Charles, D., Morrow, P., Crosbie,

J., and McDonough, S. (2010). Designing engag-

ing, playable games for rehabilitation. In Proc. 8th

Intl Conf. on Disability, Virtual Reality and Assoc.

Technologies, Vi

na del Mar/Valpara

ıso, Chile, August,

pages 195–201.

Calderita, L., Bachiller, P., Bandera, J., and Bustos, P.

(2011). Mimic: A human motion imitation component

for robocomp. In Proceedings of the 1st Workshop on

Recognition and Action for Scene Understanding (RE-

ACTS 2011), M

alaga, Spain.

Calderita, L. V., Bandera, J. P., Bustos, P., and Skiadopou-

los, A. (2013). Model-based reinforcement of kinect

depth data for human motion capture applications.

Sensors, 13(7):8835–8855.

Cid, F., Cintas, R., Manso, L., Calderita, L., S

anchez, A.,

and N

nez, P. (2011). A real-time synchronization al-

gorithm between text-to-speech (tts) system and robot

mouth for social robotic applications. In Proceedings,

Workshop en Agentes F

ısicos (WAF 2011), Albacete,

Spain, pages 81–86.

Colombo, R., Pisano, F., Mazzone, A., Delconte, C.,

Micera, S., Carrozza, M., Dario, P., and Minuco, G.

(2007). Design strategies to improve patient motiva-

tion during robot-aided rehabilitation. Journal of Neu-

roEngineering and Rehabilitation, 4(1):1–12.

Fasola, J. and Mataric, M. (2011). Comparing physical and

virtual embodiment in a socially assistive robot exer-

cise coach for the elderly. Technical report, Technical

Report CRES-11-003, Center Robotics and Embed-

ded Systems, University of Southern California.

Feil-seifer, D. and Matari

c, M. J. (2011). Ethical principles

for socially assistive robotics.

Hermano and Hogan, N. (2009). A working model of

stroke recovery from rehabilitation robotics practi-

tioners. Journal of NeuroEngineering and Rehabili-

tation.

Jezernik, S., Colombo, G., Keller, T., Frueh, H., and Morari,

RehabilitationforChildrenwhilePlayingwithaRoboticAssistantinaSeriousGame

M. (2003). Robotic orthosis lokomat: A rehabilitation

and research tool. Neuromodulation: Technology at

the Neural Interface, 6(2):108–115.

Khoshelham, K. and Elberink, S. O. (2012). Accuracy and

resolution of kinect depth data for indoor mapping ap-

plications. Sensors, 12(2):1437–1454.

Leocani, L. and Comi, G. (2006). Electrophysiological

studies of brain plasticity of the motor system. Neuro-

logical Sciences, (1).

Manso, L., Bachiller, P., Bustos, P., N

nez, P., Cintas, R.,

and Calderita, L. (2010). Robocomp: a tool-based

robotics framework. In Simulation, Modeling, and

Programming for Autonomous Robots, pages 251–

262. Springer.

Mej

ıas, C. S., Echevarr

ıa, C., Nu

nez, P., Manso, L., Bus-

tos, P., Leal, S., and Parra, C. (2013). Ursus: A

robotic assistant for training of children with motor

impairments. In Converging Clinical and Engineer-

ing Research on Neurorehabilitation, pages 249–253.

Springer.

Michelow, B., Clarke, H., Curtis, C., Zuker, R., Seifu, Y.,

and Andrews, D. (1994). The natural history of ob-

stetrical brachial plexus palsy. Plast. Reconstr. Surg.,

93(4):675–680.

Rego, P., Moreira, P. M., and Reis, L. P. (2010). Seri-

ous games for rehabilitation: A survey and a classi-

ﬁcation towards a taxonomy. In Information Systems

and Technologies (CISTI), 2010 5th Iberian Confer-

ence on, pages 1–6. IEEE.

Steultjens, E. M., Dekker, J., Bouter, L. M., van de Nes,

J. C., Cup, E. H., and van den Ende, C. H. (2003).

Occupational therapy for stroke patients a systematic

review. Stroke, 34(3):676–687.

Tapus, A., Mataric, M. J., and Scassellati, B. (2007). So-

cially assistive robotics. IEEE Robotics and Automa-

tion Magazine, 14(1):35.

Timmermans, A. A., Seelen, H. A., Willmann, R. D., and

Kingma, H. (2009). Technology-assisted training of

arm-hand skills in stroke: concepts on reacquisition

of motor control and therapist guidelines for rehabil-

itation technology design. Journal of neuroengineer-

ing and rehabilitation, 6:1.

Whitall, J., Waller, S. M., Silver, K. H. C., and Macko,

R. F. (2000). Repetitive Bilateral Arm Training With

Rhythmic Auditory Cueing Improves Motor Function

in Chronic Hemiparetic Stroke. Stroke, (10).

NEUROTECHNIX2013-InternationalCongressonNeurotechnology,ElectronicsandInformatics