Data Augmentation of Wearable Sensor Data for Parkinson’s
Disease Monitoring using Convolutional Neural Networks
∗
Terry T. Um
University of Waterloo
Canada
terry.t.um@gmail.com
Franz M. J. Pster
Ludwig-Maximilians-Univ. München
Germany
fmj.pster@me.com
Daniel Pichler
Technical University of Munich
Germany
daniel.pichler@tum.de
Satoshi Endo, Muriel Lang,
Sandra Hirche
Technical University of Munich
Germany
{s.endo,muriel.lang,hirche}@tum.de
Urban Fietzek
Schön Klinik München Schwabing
Germany
urban.etzek@schoen-kliniken.de
Dana Kulić
University of Waterloo
Canada
dana.kulic@uwaterloo.ca
ABSTRACT
While convolutional neural networks (CNNs) have been success-
fully applied to many challenging classication applications, they
typically require large datasets for training. When the availability
of labeled data is limited, data augmentation is a critical prepro-
cessing step for CNNs. However, data augmentation for wearable
sensor data has not b een deeply investigated yet.
In this paper, various data augmentation methods for wearable
sensor data are proposed. The proposed methods and CNNs are
applied to the classication of the motor state of Parkinson’s Dis-
ease patients, which is challenging due to small dataset size, noisy
labels, and large intra-class variability. Appropriate augmentation
improves the classication performance from 77.54% to 86.88%.
CCS CONCEPTS
· Computing methodologies → Supervised learning by clas-
sication; · Applied computing → Consumer health;
KEYWORDS
Data augmentation; wearable sensor; convolutional neural net-
works; Parkinson’s disease; health monitoring
ACM Reference Format:
Terry T. Um, Franz M. J. Pster, Daniel Pichler, Satoshi Endo, Muriel Lang,
Sandra Hirche, Urban Fietzek, and Dana Kulić. 2017. Data Augmentation of
Wearable Sensor Data for Parkinson’s Disease Monitoring using Convolu-
tional Neural Networks. In Proceedings of 19th ACM International Conference
on Multimodal Interaction (ICMI’17). ACM, New York, NY, USA, 5 pages.
https://doi.org/10.1145/3136755.3136817
∗
This work was partly supported by the EU Seventh Framework Programme FP7/2007-
2013ithin the ERC Starting Grant Control based on Human Models (con-humo), grant
agreement No. 337654.
Permission to make digital or hard copies of all or part of this work for p ersonal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specic permission and/or a
fee. Request permissions from permissions@acm.org.
ICMI’17, November 13ś17, 2017, Glasgow, UK
© 2017 Association for Computing Machinery.
ACM ISBN 978-1-4503-5543-8/17/11. . . $15.00
https://doi.org/10.1145/3136755.3136817
1 INTRODUCTION
In recent years, convolutional neural networks (CNNs) have shown
excellent performance on classication problems when large-scale
labeled datasets are available (e.g. [
9
,
20
]). However, it is challenging
to apply CNNs to problems where only small labelled datasets are
available. For example, collecting and labeling a large amount of
medical data is often dicult. As a result, it is challenging to apply
CNNs to small-scale medical data.
Data augmentation leverages limited data by transforming the
existing samples to create new ones. A key challenge for data aug-
mentation is to generate new data that maintains the correct label,
which typically requires domain knowledge. However, it is not
obvious how to carry out label-preserving augmentation in some
domains, e.g., wearable sensor data. For example, scaling of the
acceleration data may change their labels because some labels are
dierentiated by the intensity of motion.
In this paper, the problem of classifying the motor state of Parkin-
son’s disease (PD) patients is tackled using CNNs. PD motor state
classication is a challenging task due to noisy labels, irrelevant
motion interference, large variability over patients, and limited
availability of the labelled data. In this paper, we propose data aug-
mentation methods for wearable sensor data and successfully tackle
the challenging PD classication task using CNNs.
The contributions of the paper can be summarized as follows:
•
Application of CNNs to the task of PD motor state classica-
tion, using a clinician-labeled dataset of 30 PD patients (25
patient’s data are exploited) in daily-living conditions.
•
A set of approaches for data augmentation of wearable sensor
datasets for CNN-based classication.
•
Experimental comparison of proposed data augmentation
methods.
2 RELATED WORK
Most PD patients experience motor uctuations, which are charac-
terized by phases of bradykinesia, i.e. underscaled and slow move-
ment, and dyskinesia, i.e. overowing spontaneous movement [
17
].
Dopaminergic treatment can alleviate symptoms of bradykinesia
while its over-treatment can cause dyskinesia. Thus, an accurate
evaluation of a patient’s phenomenology is needed for determining
the right dose of medication. Current PD motor state evaluation
216
ICMI’17, November 13ś17, 2017, Glasgow, UK T. Um, F. Pfister, D. Pichler, S. Endo, M. Lang, S. Hirche, U. Fietzek, and D. Kulić
(a)Bardykinesia
(typical)
(b)Dyskinesia
(typical)
(c)Bardykinesia
(atypical)
(d)Dyskinesia
(atypical)
Figure 1: (a) and (b) show typical examples of bradykinesia
and dyskinesia in a 1 min window while (c) and (d) show
atypical patterns. The blue, red, green represent X,Y,Z sig-
nals from the accelerometer, respectively.
methods rely on patient self-reports and visual observation by the
clinician [17].
Researchers have proposed automating the evaluation with wear-
able sensors (e.g. [
5
,
19
]). However, most approaches to date have
been limited to standardized motor tasks in clinical settings [
18
].
To enable automated evaluation of PD motor states which covers
a wide range of PD symptoms across patients, a large amount of
wearable sensor data in daily-living conditions is needed [
4
]. Deep
learning (DL) approaches [
15
] provide a promising methodology
to deal with the large variability of PD data [
5
,
8
,
13
]. Given the
diculty in collecting such large datasets, data augmentation is
needed [6].
Data augmentation is an indisp ensable preprocessing step for
achieving peak performance in DL approaches (e.g. [
9
,
12
]). For
augmenting time-series data, Le Guennec et al. [
14
] used window
slicing and window warping methods, which extracts multiple
small-size windows from a single window and lengthens/shortens a
part of the window data, respectively. Unlike data augmentations for
image [
2
] and speech recognition [
3
], however, data augmentation
for wearable sensor data has not b een systematically investigated
yet to the best of our knowledge. In this paper, we propose various
data augmentation methods that enable the classication of PD
motor states from wearable data and evaluate them using CNN.
3 PD MOTOR STATE CLASSIFICATION
3.1 Challenges in PD Data
We consider two frequent PD motor states: bradykinesia, which
is characterize d by decreased movement speed and may be ac-
companied by tremor, and dyskinesia, which is characterized by
involuntary extremity movements. Figure
1 illustrates exemplar
one minute data windows of both motor states, from a single ac-
celerometer worn on the wrist of PD patients. Bradykinesia data
typically appear as constant signals indicating less movement (Fig
1(a)) while dyskinesia data consist of uctuating movements (Fig
1(b)).
However, there are a signicant number of examples that devi-
ate from the stereotypical expressions. For example, bradykinesia
accompanied by tremor can show uctuating signals which look
like a dyskinesia state (Fig 1(c)). On the other hand, dyskinesia with
voluntary suppression can show constant signals which look like a
bradykinesia state (Fig 1(d)).
There are several factors that can cause an apparent disagree-
ment between the obser ved data pattern and the expert label. First,
if the body of the patient indicates, e.g., a dyskinesia state, but the
hand which wears the wearable sensor does not move because the
patient is, e.g., holding a chair for suppressing the symptom, the
assigned label based on the overall body expression will be mis-
matched with the recorded data from the wearable device. Also,
the expert rater typically rates the symptoms for a xe d length
window, but arbitrary segmentation into xed length windows
may not result in single motor state windows. Furthermore, the
interference of voluntary movements, e.g., waving the hand, can
make bradykinesia states look like dyskinesia, and, e.g, voluntary
rest, appear like bradykinesia. Finally, bradykinesia accompanied
by tremor can also can make it dicult to distinguish between
bradykinesia and dyskinesia.
The factors described above introduce noisy labels, and lead to
large intra-class variability and signicant overlap between two
classes. As a result, it makes the PD motor state classication more
challenging, particularly given a small amount of data.
3.2 Data Augmentation Methods for Wearable
Sensor Data
Data augmentation can be viewed as an injection of prior knowl-
edge about the invariant properties of the data against certain trans-
formations. Augmented data can cover unexplored input space,
prevent overtting, and improve the generalization ability of a
DL model [
6
]. In image recognition, it is well-known that minor
changes due to jittering, scaling, cropping, warping and rotating do
not alter the data labels because they are likely to happen in real
world observations. However, label-preserving transformations for
wearable sensor data are not obvious and intuitively recognizable
(Fig 2).
One factor that can introduce label-invariant variability of wear-
able sensor data are dierences in sensor placement between partic-
ipants. For example, an upside-down placement of the sensor can
invert the sign of the sensor readings without changing the labels.
Therefore, augmentation by applying arbitrary
rotations (Rot)
to
the existing data can be used as a way of simulating dierent sensor
placements.
Another factor that can introduce variability is the temporal
location of activity events, e.g., tremor, in the window. Since the
xed size window segmentation is arbitrary, the location of the
observed symptom in the window does not have any meaning. Thus,
we may augment data by perturbing the location of the windows
or events.
Permutation (Perm)
is a simple way to randomly perturb the
temporal location of within-window events. To perturb the location
of the data in a single window, we rst slice the data into
N
same-
length segments, with N ranging from 1 to 5, and randomly permute
the segments to create a new window.
Time-warping (TimeW)
is another way to perturb the temporal location. By smoothly dis-
torting the time intervals between samples, the temporal locations
of the samples can b e changed using time-warping.
Small changes in magnitude may preserve the labels, dep end-
ing on the target task.
Scaling (Scale)
changes the magnitude of
the data in a window by multiplying by a random scalar, while
217
Data Augmentation of Wearable Sensor Data for Parkinson’s Disease Monitoring ... ICMI’17, November 13ś17, 2017, Glasgow, UK
No Aug Jittering Scaling Rotation Permutation MagWarp TimeWarp Cropping
Rot+Perm1 Rot+Perm2 Rot+Perm3 Rot+Perm4 MagW+TimeW1 MagW+TimeW2 MagW+TimeW3 MagW+TimeW4
Figure 2: Various data augmentations that are used in the experiments: jittering, scaling, rotating, permutating, magnitude-
warping, time-warping methods. Combinations of various data augmentations can also be applied.
magnitude-warping (MagW)
changes the magnitude of each
sample by convolving the data window with a smooth curve vary-
ing around one. In addition,
jittering (Jitter)
is also considered as
a way of simulating additive sensor noise. These data augmenta-
tion methods may increase robustness against multiplicative and
additive noise and improve performance.
Lastly,
cropping (Crop)
, which is similar to image cropping or
window slicing in [
14
], is applied for diminishing the dependency
on event locations. Note that cropping can capture an event-free
region, which might change the label. Also, note that cropping with
random locations over epochs will eventually converge to a sliding
window metho d with arbitrary stride sizes.
In a nutshell, jittering, scaling, cropping, rotating, permutating,
magnitude-warping and time-warping methods are applied for aug-
menting wearable sensor data. In the next section, the performance
of PD motor state classication with the proposed data augmenta-
tion methods is evaluated using CNNs.
4 EXPERIMENTS
4.1 Data Preparation
A dataset of 30 patients’ motor states was collected using Microsoft
Band 2 [
1
] in daily-living conditions without requesting specic
motor tasks
1
. The 30 PD patients are 67
±
10 years old, median
Hoehn & Yahr stage 2, average disease duration 11
±
5 years, and
MoCA points 26
±
3. Among them, 25 patient’s data are used for this
research and each one minute interval is labeled by a clinical expert.
The data are collecte d at a frequency of 62.5Hz and resampled to
120Hz to deal with sampling irregularities. The rst 58-seconds of
data (6960 samples) from each one minute window is used to make
same-length instances.
Similar to previous works (e.g. [
19
], [
7
], [
8
], [
5
]) acceleration
data only are used for the PD motor state classication. Also, no-
symptom data are removed to simplify the problem and focus on
characterizing data augmentation methods. Data collected during
walking, laying and eating activities are also removed due to limited
observation of movement during these activities. Note that no other
preprocessing, e.g., data normalization or smoothing, is applied
because they may confound the data label and subsequent results.
The resulting dataset consists of 3530 min (58.8 hours) of bradyki-
nesia and dyskinesia data. For cross-validation, the 25 PD patients
1
The study was approved by the ethics committee of Technical University of Munich
(Az. 234/16 S).
⋮
⋮
⋮
⋮
⋮
⋮
6960*3 2319*3 772*3 385*3 193*3 97*3 49*3 48*1 48 2
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
input conv1(16) conv2(32) conv3(64) conv4(64) conv5(64) conv6(64) conv7(64) GAP(1) output
input size
Figure 3: 7-layer CNN with a global average pooling (GAP).
The 7-layer CNN consists of 16-32-64-64-64-64-64 feature
maps which reduce the size of the inputs to 2319*3, 772*3,
385*3, 193*3, 97*3, 49*3, 48*1, respectively.
are divided into ve subject groups. The performance of PD mo-
tor state classication is reported in Section 4.3 using the average
values of 5-fold cross-validation results.
4.2 The CNN Architecture
In this research, CNNs are used for PD motor state classication.
CNNs are more suitable for small-scale datasets than long short-
term memories (LSTMs) [
10
] because CNNs generally use a smaller
number of parameters compared to fully-connected LSTMs. Deep
and sparse 7-layer CNNs (Figure 3) are employed to capture the
large variability of the small-scale PD data.
A convolutional layer, a batch normalization layer [
11
], and an
activation layer using rectied units (ReLUs) form a single con-
volutional layer of the 7-layer CNN. With strided convolutions
using 4*1, 4*1, 3*1, 3*3, 2*3, 2*3, 2*3 convolution lters, the sizes
of the inputs are reduced from 6960*3 to 48*1 over layers (Figure
3). Note that XYZ signals of the accelerometer are convolved in
layers 4,5,6 and 7 to capture inter-vector-component features. For
reducing the number of parameters for small-scale datasets, a global
averaging pooling (GAP) layer [
16
] is applied at the end instead of
fully-connected layers.
4.3 Results
Classication of PD motor states is performed using the CNN with
the various data augmentation methods. For baseline results, a sup-
port vector machine (SVM) with an RBF kernel is applied to 540
dimensional statistical features: mean, variance, skewness, kurtosis,
218
ICMI’17, November 13ś17, 2017, Glasgow, UK T. Um, F. Pfister, D. Pichler, S. Endo, M. Lang, S. Hirche, U. Fietzek, and D. Kulić
Table 1: The results of PD motor state classication with
various data augmentation methods. R,P,T,M represent Rot,
Perm, TimeW, MagW, respectively.
SVM CNN Jitter Scale Crop Rot Perm
Train 98.82 99.92 99.78 99.84 65.77 100.0 99.33
Test 70.72 77.54 77.52 79.46 73.58 82.62 81.16
MagW TimeW P,T R,P R,T R,P,T R,P,T,M
Train 100.0 94.67 96.63 99.08 94.70 94.43 94.20
Test 79.33 82.00 81.75 86.76 85.01 86.88 85.60
0 100 200 300 400
Epoch
30
40
50
60
70
80
90
100
Accuracy (%)
(a) Training Accuracy
CNN
Rot
Perm
TimeW
R,P
R,P,T
0 100 200 300 400
Epoch
30
40
50
60
70
80
90
100
Accuracy (%)
(b) Test Accuracy
CNN
Rot
Perm
TimeW
R,P
R,P,T
Figure 4: Training curves for CNN, Rot, Perm, TimeW,
Rot+Perm and Rot+Perm+TimeW methods. The curves of
Rot+Perm+TimeW shows slow training improvement and a
better generalization performance.
and maximum values are extracted from 1 min data using 5 and
10-sec sliding windows. Also, a CNN is applied to raw 1 min data
without data augmentation for baseline comparison. All experi-
ments except for the SVM are performed for 400 epochs and the
median values from the last 10 epoch results are used for averaging
the 5-fold cross-validation results.
Dierent random parameter values are applied for data augmen-
tation. For jittering, a standard deviation (STD) value is sampled
from a Gaussian distribution with 0.03 STD, and 1 min of Gaussian
noise is generated using the sampled STD value. For scaling, a ran-
dom scalar is sampled from a Gaussian distribution with a mean of 1
and 0.1 STD. For rotation, an arbitrary rotation matrix is generated
for each instance. For permutation, a random integer
N
is deter-
mined by rounding a positive value sampled from a Gaussian dis-
tribution with 5.0 STD. For magnitude-warping and time-warping,
random sinusoidal curves are generated using arbitrary amplitude,
frequency, and phase values. The implemented code for the pro-
posed data augmentation methods is available online: https://github.
com/terryum/Data-Augmentation-For-Wearable-Sensor-Data
The main results are presented in Table 1. Jittering fails to im-
prove the performance of PD motor state classication because
it introduces rapid uctuations which look similar to dyskinesia.
Cropping also fails because it drops the information of 2
/
3 win-
dow samples, which could be a critical loss given the small dataset.
Cropping of an event-free region also hinders the learning process
and can be a cause of the poor performance. Scaling and magnitude-
warping also fail be cause changing of the intensity of the signal
may alter the labels.
Figure 5: Randomly selected 40 incorrect mispredictions
from the Fold-1 results of Rot+Perm+TimeW experiment.
Fluctuating signals from bradykinesia (white) and constant
signals from dyskinesia (yellow) are often misclassied.
On the other hand, rotation, permutation, and time-warping
methods improve the performance of PD motor state classication.
The best performance among the single data augmentation methods
is achieved by rotation. Permutation and time-warping also provide
performance improvements by perturbing the temporal locations
of samples. These results indicate that the major sources of vari-
ability are dierent sensor placements between participants and
event locations in an arbitrarily segmented window. The proposed
rotation, permutation, and time-warping methods eectively com-
pensate the unnecessary variations and improve the performance
by 3.6-5.1% accuracy.
Combinations of various data augmentation methods show bet-
ter performance than that of a single data augmentation method.
The combinations of Rot+Perm and Rot+TimeW show better perfor-
mance than the baseline of CNN by 7.5-9.2%. The best performance
is achieved by Rot+Perm+TimeW with 86.88% accuracy. These re-
sults indicate that rotation can be used to alleviate sensor pose
variability while either permutation or time-warping can be em-
ployed for addressing the variability of temporal locations of events
in a window.
Training curves of the experiments are depicted in Fig 4. The
Rot+Perm+TimeW curve shows slow training improvement and a
better generalization performance than others thanks to the regu-
larization eect provided by the data augmentation. Some of the
failed predictions are presented in Fig 5. From the gure, it can be
observed that CNNs often misclassify uctuating bradykinesia and
constant dyskinesia data, which can be considered as seemingly-
noisy labels as described in Section 3.1.
5 CONCLUSION
In this paper, an automatic classication algorithm for PD motor
state monitoring is developed based on wearable sensor data. PD
motor state classication is a challenging task because of large
inter-class variability, noisy labels, interference by irrelevant mo-
tion signals and limited data availability. The challenging PD task
is successfully tackled using a 7-layer CNN and the proposed data
augmentation methods. The combination of rotational and permu-
tational data augmentation methods improves the baseline perfor-
mance of 77.52% accuracy to 86.88%. Systematic experiments with
various data augmentation methods provide a direction towards a
general approach for augmentation for wearable sensor data.
219
Data Augmentation of Wearable Sensor Data for Parkinson’s Disease Monitoring ... ICMI’17, November 13ś17, 2017, Glasgow, UK
REFERENCES
[1] Microsoft Band 2. 2015.
https://www.microsoft.com/microsoft-band/
[2]
Ken Chateld, Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2014.
Return of the Devil in the Details: Delving Deep into Convolutional Nets. CoRR
abs/1405.3531 (2014).
[3]
Xiaodong Cui, Vaibhava Goel, and Brian Kingsbury. 2015. Data Augmentation
for Deep Neural Network Acoustic Modeling. IEEE/ACM Trans. Audio, Speech
and Lang. Proc. 23, 9 (Sept. 2015), 1469ś1477.
[4]
Silvia Del Din, Alan Godfrey, Claudia Mazzà, Sue Lord, and Lynn Rochester. 2016.
Free-living monitoring of Parkinson’s disease: Lessons from the eld. Movement
Disorders 31, 9 (2016), 1293ś1313.
[5]
Bjoern M. Eskoer, Sunghoon I. Lee, Jean-Francois Daneault, Fatemeh N. Go-
labchi, Gabriela Ferreira-Carvalho, Gloria Vergara-Diaz, Stefano Sapienza, Gian-
luca Costante, Jochen Klucken, Thomas Kautz, and Paolo Bonato. 2016. Recent
machine learning advancements in sensor-based mobility analysis: Deep learning
for Parkinson’s disease assessment. In 2016 38th Annual International Conference
of the IEEE Engineering in Medicine and Biology Society (EMBC). 655ś658.
[6]
Ian Goodfellow, Yoshua Bengio, and Aaron Courville. 2016. Deep Learning. MIT
Press.
[7]
Robert I. Griths, Katya Kotschet, Sian Arfon, Zheng Ming Xu, William Johnson,
John Drago, Andrew Evans, Peter Kempster, Sanjay Raghav, and Malcolm K.
Horne. 2012. Automated assessment of bradykinesia and dyskinesia in Parkin-
son’s disease. Journal of Parkinson’s disease 2, 1 (2012), 47ś55.
[8]
Nils Y. Hammerla, James M. Fisher, Peter Andras, Lynn Rochester, Richard Walker,
and Thomas Plotz. 2015. PD Disease State Assessment in Naturalistic Environ-
ments Using Deep Learning. In Proceedings of the Twenty-Ninth AAAI Conference
on Articial Intelligence (AAAI’15). AAAI Press, 1742ś1748.
[9]
Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2015. Deep Residual
Learning for Image Recognition. CoRR abs/1512.03385 (2015). http://arxiv.org/
abs/1512.03385
[10]
Sepp Hochreiter and Jörgen Schmidhuber. 1997. Long Short-Term Memory.
Neural Computation 9, 8 (Nov 1997), 1735ś1780.
[11]
Sergey Ioe and Christian Szegedy. 2015. Batch Normalization: Accelerating
Deep Network Training by Reducing Internal Covariate Shift. ArXiv e-prints
(Feb. 2015). arXiv:cs.LG/1502.03167
[12]
Alex Krizhevsky, Ilya Sutskever, and Georey E. Hinton. 2012. ImageNet Clas-
sication with Deep Convolutional Neural Networks. In Advances in Neural
Information Processing Systems 25, F. Pereira, C. J. C. Burges, L. Bottou, and K. Q.
Weinberger (Eds.). Curran Associates, Inc., 1097ś1105.
[13]
Ken J. Kubota, Jason A. Chen, and Max A. Little. 2016. Machine learning for
large-scale wearable sensor data in Parkinson’s disease: Concepts, promises,
pitfalls, and futures. Movement Disorders 31, 9 (2016), 1314ś1326.
[14]
Arthur Le Guennec, Simon Malinowski, and Romain Tavenard. 2016. Data Aug-
mentation for Time Series Classication using Convolutional Neural Networks.
In ECML/PKDD Workshop on Advanced Analytics and Learning on Temporal Data.
Riva Del Garda, Italy. https://halshs.archives-ouvertes.fr/halshs-01357973
[15]
Yann LeCun, Yoshua Bengio, and Georey Hinton. 2015. Deep learning. Nature
521, 7553 (28 May 2015), 436ś444. Insight.
[16]
Min Lin, Qiang Chen, and Shuicheng Yan. 2013. Network In Network. ArXiv
e-prints (De c. 2013). arXiv:1312.4400
[17]
João Massano and Kailash P. Bhatia. 2012. Clinical Approach to Parkinson’s
Disease: Features, Diagnosis, and Principles of Management. Cold Spring Harb
Perspect Med 2, 6 (Jun 2012), a008870.
[18]
Christiana Ossig, Angelo Antonini, Carsten Buhmann, Joseph Classen, Ilona
Csoti, Björn Falkenburger, Michael Schwarz, Jürgen Winkler, and Alexander
Storch. 2016. Wearable sensor-based objective assessment of motor symptoms in
Parkinson’s disease. Journal of Neural Transmission 123, 1 (2016), 57ś64.
[19]
Shyamal Patel, Konrad Lorincz, Richard Hughes, Nancy Huggins, John Growdon,
David Standaert, Metin Akay, Jennifer Dy, Matt Welsh, and Paolo Bonato. 2009.
Monitoring Motor Fluctuations in Patients With Parkinson’s Disease Using
Wearable Sensors. IEEE Transactions on Information Technology in Biomedicine
13, 6 (Nov 2009), 864ś873.
[20]
Terry Taewoong Um, Vahid Babakeshizadeh, and Dana Kulić. 2017. Exercise Mo-
tion Classication from Large-Scale Wearable Sensor Data Using Convolutional
Neural Networks. In 2017 IEEE/RSJ International Conference on Intelligent Robots
and Systems (IROS). 1ś6.
220