=Paper=
{{Paper
|id=Vol-2069/STREAMEVOLV4
|storemode=property
|title=Improving Human Activity Classification through Online Semi-Supervised Learning
|pdfUrl=https://ceur-ws.org/Vol-2069/STREAMEVOLV4.pdf
|volume=Vol-2069
|authors=Hugo Cardoso,João Mendes-Moreira
|dblpUrl=https://dblp.org/rec/conf/pkdd/CardosoM16
}}
==Improving Human Activity Classification through Online Semi-Supervised Learning==
https://ceur-ws.org/Vol-2069/STREAMEVOLV4.pdf
Improving Human Activity Classification
through Online Semi-Supervised Learning
Hugo Cardoso? and João Mendes-Moreira
INESC TEC, Faculty of Engineering, University of Porto, R. Dr. Roberto Frias s/n,
4200-465 Porto, Portugal
hugo.l.cardoso@inesctec.pt; xkynar.github.io
Abstract. Built-in sensors in most modern smartphones open multiple
opportunities for novel context-aware applications. Although the Hu-
man Activity Recognition field seized such opportunity, many challenges
are yet to be addressed, such as the differences in movement by peo-
ple doing the same activities. This paper exposes empirical research on
Online Semi-supervised Learning (OSSL), an under-explored incremen-
tal approach capable of adapting the classification model to the user
by continuously updating it as data from the user’s own input signals
arrives. Ultimately, we achieved an average accuracy increase of 0.18
percentage points (PP) resulting in a 82.76% accuracy model with Naive
Bayes, 0.14 PP accuracy increase resulting in a 83.03% accuracy model
with a Democratic Ensemble, and 0.08 PP accuracy increase resulting
in a 84.63% accuracy model with a Confidence Ensemble. These models
could detect 3 stationary activities, 3 active activities, and all transitions
between the stationary activities, totaling 12 distinct activities.
Keywords: Human Activity Recognition, Machine Learning, Online
Semi-Supervised Learning
1 Introduction
The goal of Human Activity Recognition (HAR) is to develop systems capable
of recognizing the actions and goals of a human agent by automatically analyz-
ing these ongoing events and extract their context from the captured data. The
detection of human activities, such as walking, running, falling, or even cycling,
has several applications, from surveillance systems to patient monitoring sys-
tems. Despite being a particularly active field of study in the past years, HAR
still leaves many strategies left to explore and key aspects left to address.
There are two main approaches in terms of data extraction: Video and sen-
sors. The sensor approach is, however, the most promising, due to its extreme
?
This work is financed by the ERDF – European Regional Development Fund through
the Operational Programme for Competitiveness and Internationalisation - COM-
PETE 2020 Programme and by National Funds through the Portuguese funding
agency, FCT - Fundação para a Ciência e a Tecnologia within project POCI-01-
0145-FEDER-016883.
2 Hugo Cardoso, João Mendes-Moreira
portability and unobtrusiveness. In particular, the introduction of these built-in
hardware sensors in many of the modern smartphones, in association with their
viral spread throughout the world, unlocked the possibility for the creation of
applications based on the context perceived from the data they provide, in a
way so vast that it could never have been envisioned a decade ago.
Most sensor-based HAR systems are trained in a static dataset with Super-
vised Learning (SL) techniques, generating a classification model with a rela-
tively low error rate. However, these systems commonly ignore one of HAR’s
challenges, the difference of input signals produced by different people when
doing the same activities. Consequently, as a user’s movements drift from the
generic, the system error increases. In fact, each user has his own unique signal,
allowing the use of accelerometers to identify them [1]. The activity classification
method should therefore be able to generate adapted results for each different
user.
The ideal scenario for this problem would be the creation of a smartphone
application capable from the beginning of classifying the user’s activities with a
certain error, and as the time passes and the user utilizes the application, without
manual input, the classification error of the system would decrease autonomously
until it is virtually insignificant for that specific user.
This document exposes a series of practical experiments performed in an
effort to provide a solution to this problem, by using an under-explored tech-
nique named Online Semi-supervised Learning (OSSL), an incremental approach
capable of adapting the classification model to the user of the application by
continuously updating it as the data from the user’s own specific input signals
arrive.
First, a brief introduction to OSSL will be conducted. Afterwards, for each
experiment, the goals, methodology, results and conclusions will be presented
sequentially, as an effort to answer some of these undressed aspects. In the end,
conclusions about both the advantages and drawbacks of this technique will be
discussed.
2 Online Semi-Supervised Learning (OSSL)
Most common Data Mining approaches make use of static datasets. These datasets
are collected and organized a priori, and only afterwards analyzed and processed.
They also have the advantage of being traditionally labeled, i.e., the activity of
the instances used for training and testing the models are known.
Labeled data is massively used by SL techniques. These approaches analyze
the data and generate a model, capable of determining the class labels for unseen
instances. Systems which perform a single training phase on a static dataset and
whose models do not change afterwards are classified as offline.
As interesting as these concepts are, the resulting systems can only be generic.
If applied in the intended HAR application, while they might yield decent results
using static datasets, gathered from a sample population, they are far from
perfect, since as the movements of a specific user drift away from the generic,
Human Activity Recognition with Online Semi-Supervised Learning 3
the classification gets worse. Therefore, the need to explore different approaches
increases. Adaptation must be taken into account.
Yet, data coming directly from the smartphone sensors would not be labeled.
It is not possible to exploit the conventional supervised learning approaches. All
these issues might be addressed by means of Online Semi-supervised Learning.
Semi-supervised Learning has the particularity of operating on both labeled
and unlabeled data, typically a small amount of labeled versus a large amount
of unlabeled one. The fact is that it is much easier to acquire unlabeled data
than labeled one. Labeled data often requires a skilled human agent, that is, an
entity which can accurately classify the data. On the contrary, the acquisition
of unlabeled data is relatively inexpensive.
OSSL is capable of adapting a classification model, pre-trained by means of
supervised learning, to the user of the application by continuously updating it
as the data from the user’s own specific input signals arrives. It is, therefore,
a promising approach to solve the problem in question. If the HAR application
learns in an online, semi-supervised fashion, it could start off as generic, but
iteratively adapt to a new user, instead of being bound by a generic model,
taught by static, pre-gathered data.
3 Empirical Research
After exposing the main architecture of the used dataset, the experiments will
be presented in the exact order they were performed. This order was important
to tackle the final problem incrementally, adding one layer of complexity to the
solution at a time. The experiments were done using MOA - Massive Online
Analysis framework [2].
3.1 Dataset
There is a significantly limited collection of activity datasets with easy access
and availability to the public. The chosen dataset from within the unfortunately
scarce possible solutions was created by Smartlab, and consists of experiments
carried out by a group of 30 volunteers, within an age bracket of 19-48 years.
It is important to state that Smartlab made an invaluable contribution by
creating such a complete and organized dataset. The reasons that render the
dataset non-optimal are very specific to this particular project, and will be dis-
cussed later on.
Each volunteer performed a protocol composed of six basic activities: Stand-
ing, Sitting, Laying, Walking, Walking Downstairs, Walking Upstairs. The tran-
sitions between the static postures also count as activities: Stand-To-Sit, Sit-To-
Stand, Sit-To-Lie, Lie-To-Sit, Stand-To-Lie, Lie-To-Stand. As such, there are a
total of 12 activities.
The recording was performed at a constant rate of 50Hz by an accelerometer
and a gyroscope of a Samsung Galaxy S II, which is very important since the
project should be able to work with readings from a smartphone. Although the
4 Hugo Cardoso, João Mendes-Moreira
dataset comes with a processed version, only the unprocessed version composed
of the original raw inertial signals from the smartphone sensors was utilized, since
it provided more freedom of choice for custom features and data summarization.
Each person repeated the activity routine twice, totaling around 15 minutes
of data per person. As such, the full dataset possess a few hours of activities,
which ended up being less than the desired, but enough to achieve results.
3.2 Perfect Segmentation Online Supervised Learning (OSL)
Goals Perfect Segmentation is what we call knowing exactly where an activity
starts and where it ends. Perfect Segmentation is virtually impossible in a real-
istic context, because it is impossible to separate a stream of sensor data into
their respective activities automatically. If that was possible, there would be no
need for human agents to spend so much time segmenting the collected data for
labeling.
However, the goal of this first experiment was not to be realistic, but to serve
as a test bed for finding the features that better identify the activity data. Also,
it was a good way of finding out the maximum accuracy that could be expected,
since it was very unlikely that higher accuracy could be achieved in a realistic
context.
Methodology This very simple initial experiment consisted of splitting the
dataset by a split factor (usually 50%), creating a train set and a test set, but
taking into account that each set must have a fair amount of each activity, to
avoid a training set to have no samples of a particular activity. Then, these sets
would be pre-processed into ARFF files and fed to a classifier.
Results Numerous combinations of features were experimented, mostly inspired
in already existing HAR Studies [3]. The best result was a combination of 20
features, 10 from the accelerometer and 10 from the gyroscope: 1) Arithmetic
Mean of the X, Y and Z axis of both sensors, individually and also together,
resulting in 4 features for each of the sensors; 2) Standard Deviation of the
X, Y and Z axis of both sensors, individually, resulting in 3 features for each of
the sensors. 3) Pearson Correlation of axis X and Y, Y and Z, and X and Z,
for both sensors, resulting in 3 features for each of the sensors. The results are
shown in Table 1.
Table 1. Results for Perfect Segmentation OSL with the final features
Naive Bayes VFDT KNN
Accuracy 87.5 % 87.5% 100%
Human Activity Recognition with Online Semi-Supervised Learning 5
Conclusions Although incremental algorithms are usually not as powerful as
their batch implementations, this experiment proves that even with just SL, it
would be possible to achieve positive results if the activities were well separated.
Although it is not a realistic situation, these features have demonstrated to
characterize the data quite well.
3.3 Perfect Segmentation Online Semi-Supervised Learning (OSSL)
Goals After finding a good set of features, it was necessary to understand
what to expect in terms of accuracy boost with the Semi-Supervised approach.
Therefore, this second experiment served as an initial familiarization with this
technique, in terms of both results and implementation. Perfect Segmentation
was still used, since the experiment does not aim at realism, but at testing limits
and data dynamics.
Methodology The first implemented OSSL algorithm was a combination of
Democratic Co-Learning [4] and Tri-training [5]. As such, a new type of classifier
was produced, an ensemble of three classifiers: Naive Bayes, VFDT and KNN.
After the initial supervised training, further training with an unseen instance
was only performed if most classifiers agreed on the label they would classify the
instance with. Since our ensemble had three classifiers, if at least two agreed,
than that new instance would be used with the agreed label for training all the
classifiers. If all disagreed, the instance would be discarded.
The testing was done by means of Leave-One-Out Cross-Validation, with the
particularity that what is left out is not an instance, but a whole user. The
reasoning behind this is due to the fact that an important goal of this entire
project is to make the application adapt to a specific user not used for training.
Assuming that the model is trained with a sample as much representative of
the population as possible, the data of 29 users are used to train the model and
one user is used to test it in a hold-out fashion.
Since the same data cannot be used simultaneously to test the initial super-
vised classifiers and to train again all the classifiers, the data of the user left
out can be: 1) split into two parts, one of them for testing the supervised model
and the second one to test the semi-supervised approach, or 2) used according
to Prequential Evaluation [6].
In this particular experiment, the data from the user left out was split, which
might not have been optimal, but it was good enough to take the desired con-
clusions. However, later experiments made use of Prequential Evaluation.
Results With this experiment, we can see in Table 2 the Democratic Ensemble
Classifier was capable of achieving a Cross-Validation accuracy average of 89.15%
before the unlabeled data was presented, and 89.49% after the unlabeled data,
that is, an increase of 0.34 percentage points (PP).
6 Hugo Cardoso, João Mendes-Moreira
Table 2. Results for Perfect Segmentation OSSL
OSL OSSL OSSL − OSL
Democratic 89.15% 89.49% 0.34 PP
Conclusions Taking into account the fact that a single individual’s data is only
around 15 minutes long, and the data was split in half for train and test, the
results are actually motivating.
This experiment proves that OSSL can indeed be used to improve a model’s
accuracy. As such, it is now time to start looking at the data as a stream.
3.4 Fixed-Length Sliding Window Online Supervised Learning
(OSL)
Goals Now that we have a general idea of what to expect in terms of results,
it was time to start working on realistic case scenarios. In this experiment, the
data is handled as a stream. As such, a fixed-length sliding window was used to
iterate the data in the order it was recorded.
Methodology The window size was set to 200, due to not only showing up
in several of the past HAR works, but because it indeed presented the most
consistent results. Since the data was recorded at a frequency of 50Hz, this
means that a window possesses 4 seconds of data.
Both non-overlapping and overlapping sliding windows were tested, with the
overlapping having an overlap value of 70%.
Since there is no Perfect Segmentation now, a method of testing the clas-
sification accuracy of a model on a window had to be decided. Initially, the
activity most present was defined as the label of the window. This meant that if
a window has 90% ”Walking” and 10% ”Standing”, the window should represent
”Walking”.
However, it turned out to be intolerant, since if an activity is just slightly less
present in a window, for example 49%, it should still be a valid classification. As
such, an updated method of validation stated that an activity should be a valid
label if it appeared in a significant amount, such as at least 30%.
Still, in the end, after some deliberation, we defined a classification as correct
if the predicted class is present in the window, since it makes sense in a realistic,
practical context.
Other than the already employed classifiers, a variant of the previous ensem-
ble classifier was created: Confidence Ensemble Classifier.
In MOA, every classifier has the getVotesForInstance method, which returns
a list with all the votes the classifier assigned to each activity. As such, for each
classifier in the ensemble, the votes were collected, equally scaled, and added. In
the end, the final classification is, simply, the most voted activity.
This differs from the Democratic Ensemble Classifier because if a classifier is
99% certain of an activity label, but the other two agree on a different activity
Human Activity Recognition with Online Semi-Supervised Learning 7
with just 30% certainty, then it is not absurd to infer that the first classifier’s
opinion should be taken into account, despite its numeric disadvantage. Also,
this classifier always presents a classification, as opposite to the Democratic,
which does not provide a classification when all the classifiers disagree.
Results Several cases were tested in this experiment. Both ensembles and their
individual classifiers were tested in Non-overlapping and Overlapping Windows.
The results are presented in Table 3.
Table 3. Results for Fixed-Length Sliding Window OSL
Naive Bayes VFDT KNN Democratic Confidence
Non-Overlapping 81.22% 81.22% 82.64% 81.22% 82.05%
Overlapping 82.11% 83.33% 82.51% 84.33% 85.41%
Conclusions From the analysis of the obtained results, we can conclude that
Overlapping Windows are significant and consistently better than Non-Overlapping
Windows.
Another interesting observation is how the ensemble classifiers are being able
to provide very competitive results in contrast to the individual classifiers. The
best result was indeed obtained from the new Confidence Ensemble Classifier,
with an accuracy of 85.41%. As this is now a realistic context, it is a satisfying
result, and a good basis for the OSSL approach.
3.5 Dynamic Data Segmentation
Goals This experiment deviated from the path that the previous experiments
were taking. In this case, an attempt at discarding fixed-length windows was per-
formed by implementing the Dynamic Data Segmentation algorithm proposed
in an article [7] by Kozina et al. Since 100% accuracy was achieved when the ac-
tivities were perfectly segmented, trying to more accurately separate the stream
of data was a logical and worthy effort.
Methodology Data was segmented when a descending acceleration peak higher
than a continuously calculated threshold was found:
threshold = (avgmax − avgmin ) × C
where C is a constant used to mitigate the impact of noise in the data. It
was firstly calculated with the method proposed in the article. However, since it
was not providing very good results, the program tested several combinations of
N (number of previous data points used to calculate the threshold) and C, in
order to find the best possible values.
8 Hugo Cardoso, João Mendes-Moreira
Results Unfortunately, the results were far worse than expected. As we can
see on Table 4, Using the same features than up until now, with the Confidence
Ensemble Classifier, an accuracy of only 67.2% was achieved. In an attempt of
optimizing the results, we experimented with the same features used in the cited
article. Although the results were better, the maximum accuracy achieved was
of 75.1%, which was still much lower than desired.
Table 4. Results for Dynamic Data Segmentation
Our features Article features
Confidence 67.2% 75.1%
Conclusions There are many reasons why this approach might not have worked
as intended. The resulting windows presented many inconsistencies. For instance,
an activity like ”Standing” could be split into notably varying intervals, rang-
ing from 8 data samples up to 50. The summarization of these windows would
therefore result in unpatterned metrics, which are prone to confusing the learning
algorithms.
Also, because the algorithm is based on finding sudden descending pikes
of acceleration, many situations in which a segment possesses two activities in
almost the same quantities happen. The reason is, for instance, that the change
between two activities is smooth enough to not be segmented into two windows
by the algorithm, failing its main purpose.
It is also hard to define what is the ideal scenario about this kind of dynamic-
length sliding window algorithm. If the segmentation is done only after the end
of an activity and the start of another, in a real case scenario, this would mean
we would only be able to process the window and know which activity we have
been doing once we actually have finished it and moved onto the next one. This
is an unacceptable user experience. If, however, the goal was to segment each
activity in every acceleration drop, we would likely acquire windows too small
to provide quality features.
Still, this is mostly speculation, and we actually believe that Dynamic Data
Segmentation might have a strong role in solving the HAR challenge. The delayed
response could always be fixed by presenting an estimate of the activity being
done after the window has a minimum size. In the end, user experience can
always be tweaked into feeling right, so it is always worth to further explore
this promising approach. Still, since it did not provide satisfying results in our
experiments, we embraced Overlapping Fixed-Length Sliding Windows for the
remaining of the project.
Human Activity Recognition with Online Semi-Supervised Learning 9
3.6 Fixed-Length Sliding Window OSSL
Goals After all the previous checkpoint experiments, we are now finally ready
to tackle head-on the concept of applying OSSL to improve the accuracy of a
generic model.
As such, the goal of this experiment was, fundamentally, the goal of the entire
project: to understand whether it is possible to use unlabeled data, acquired from
the application’s final user, to improve the generic model which composes the
initial state of the application.
Methodology As in the previous experiments, a sliding window of size 200
with 70% overlapping was used. Testing was performed with Leave-One-Out
Cross-Validation, with Prequential Evaluation. This means that each window of
data from the user that was left out would first be used for testing (the model
would try to classify the window), and only then for training, if the window was
considered a good training sample. This technique allowed to use the data to its
full potential.
One of the biggest obstacles that were faced was that several times, a window
that was considered a good training sample was labeled incorrectly, which meant
the model was being wrongfully trained.
To avoid these mistrainings, we resorted to very high thresholds of certainty.
As such, a new instance was only used for further training if every classifier
composing the ensemble had at least 99.9% of their votes in the same activity.
It might seem too restrictive, but in the end, it was more desirable to discard
an instance than to use it it wrongly.
This instance validation method also allow us to use independently the clas-
sifiers composing our ensembles, since each classifier is capable of voting. There-
fore, although each classifier and ensemble has their own ways of classifying
an instance, in this experiment they all used the same method to determine
trainable instances.
Results Table 5 shows the results of every test performed in this experiment.
For each classifier or ensemble, the model accuracy was tested before and after
the application of the unlabeled data by the Semi-Supervised methods. The table
also contains a column with the difference between the pos and pre accuracies,
for easier interpretation.
Table 5. Results for Fixed-Length Sliding Window OSSL
Naive Bayes VFDT KNN Democratic Confidence
OSL 82.58% 83.56% 76.29% 82.89% 84.55%
OSSL 82.76% 83.47% 72.72% 83.03% 84.63%
OSSL - OSL 0.18 PP -0.09 PP -3.57 PP 0.14 PP 0.08 PP
10 Hugo Cardoso, João Mendes-Moreira
As can be seen from the analysis of the result table, the ensembles remain
as the classifiers with the most consistent results. The Democratic Ensemble
was capable of achieving an average improvement of 0.14 percentage points, and
despite the Confidence Ensemble providing a lower average improvement (0.08
percentage points), its final accuracy is the highest (84.63%). Naive Bayes was
also able to achieve positive results even independently, with an average accuracy
gain of 0.18 percentage points for a final accuracy of 82.76%. VFDT and KNN
behaved not so well as independent classifiers. In average, the SSL approach
reduced the models performance. However, when working as an ensemble, their
view of the data was beneficial at achieving consistently positive results.
Conclusions This experiment was very gratifying as it proves that OSSL can
be used in practice to improve the accuracy of a model, achieving a better HAR
system.
The small amount of data per person (15 minutes) is a strong obstacle that
might explain the low increase of performance. However, increases were achieved
with nothing but some unlabeled data.
As such, it is indeed possible that, by simply processing a stream of unlabeled
data extracted from smartphone sensors, a generic model improves by itself, and
classifications that were once incorrect become accurate.
3.7 OSSL Using The Author’s Data
Goals With the desire to take one step further into proving that OSSL has
a realistic place in achieving accurate and autonomous HAR systems, the first
author himself decided to record some of his own data in similar conditions to
that of the original dataset.
The goal of this experiment was to prove that a generic model, fully trained
in a dataset built in lab conditions, can improve its performance and adapt to
anyone, even to the first author of this paper.
Methodology With the help of a waist belt, a Smartphone Galaxy S3 and a
stopwatch (for ease of labeling in order to find the accuracy improvements), the
author recorded himself performing a routine containing the same activities used
in this project:
Standing –> Sitting –> Laying –> Sitting –> Standing –> Laying –>
Standing –> Walking –> Walking Upstairs –> Walking Downstairs
The routine was repeated three times, producing a total of roughly 50 minutes
of data. Classifiers were trained in a supervised fashion in the entirety of the
original dataset, resulting in generic models. These models were tested in the
user’s data to acquire an initial accuracy value. Then, the models employed the
previous Semi-Supervised techniques to further train using the new, recorded
data. Prequential Evaluation was used in order to acquire understandable metrics
and make the most out of the author’s data.
Human Activity Recognition with Online Semi-Supervised Learning 11
Results Due to the fact the original dataset and the author’s data were col-
lected in unequal conditions, especially in terms of the waist belt used, some
inconsistencies in the data patterns were produced. These inconsistencies tended
to confuse most classifiers, and because of that, the threshold for an instance
to be considered training material had to be increased from 99.9% to 99.99%.
With this certainty, Naive Bayes was the classifier which better adapted to these
inconsistencies, producing an accuracy increase from 60.17% to 60.25%.
100% certainty threshold was also tested for this experiment, with the Confi-
dence Ensemble Classifier providing an accuracy boost from 62.18% to 62.22%.
However, in the end, all classifiers presented very low overall accuracies.
Table 6. Results for Fixed-Length Sliding Window OSSL with the author’s data
OSL OSSL OSSL − OSL
Naive Bayes (99.99% thld) 60.17 % 60.25% 0.08 PP
Confidence (100% thld) 62.18 % 62.22% 0.04 PP
Conclusions We can conclude from this experiment that even small variations
in the data gathering conditions tend to affect the accuracy of the model. Most
results from this experiment were adverse, due to the fact that a good generic
model is essential as a base for the model to be capable of training itself. The
role of OSSL in these systems should not be to turn a bad classifier into a good
one, but to turn a good classifier into a better one.
Despite the results, OSSL was still capable of improving the accuracy of
some models. It is very interesting to think that the 40 minutes of activities
the author has performed were able to help a classification model to improve its
own accuracy, even if only by a little. This experiment was very important to
understand both the role and the limitations of OSSL.
4 Conclusion and future work
The empirical research was capable of demonstrating that OSSL may indeed
provide improvements to a generic model, adapting it to a specific user. Accuracy
gains of up to 0.18 PP were achieved with just 15 minutes of unlabeled data.
It was also concluded that OSSL works better when the base generic model
has a good initial accuracy, since a competent classifier is more qualified for
self-training. Therefore, it excels at making good classifiers even better.
However, a lot of research is still essential to even think about turning this
technology into an everyday tool. In terms of the chosen dataset, its main inad-
equacy to the project was due to the fact that the project focused on proving
that OSSL improved a generic model for a specific user. This means that while
the generic model can be trained from the data of several individuals, the data
12 Hugo Cardoso, João Mendes-Moreira
used for testing and Semi-Supervised training must come from a single user. As
such, it would be important to have recordings for each individual longer than
15 minutes.
The dataset was also recorded using a waist belt, which ends up being very
obtrusive. It would be much more realistic whether the new dataset had the
smartphone located in the user’s front pocket. However, that approach should be
further researched because orientation matters. Very different data is produced
depending on the smartphone orientation even when inside the pocket, which will
likely confuse the classification model and render the application useless. This
is unacceptable in terms of user experience. As such, a method for converting
sensor data from the smartphone orientation to a generic orientation should be
developed. The paper [9] is an attempt to solve this issue, and it should serve as
a basis for additional attempts, especially applied to HAR and OSSL.
The addressing of the stated considerations may or may not be enough to
solve the massive challenge that is Human Activity Recognition, but we believe
they are key steps in turning this technology into an everyday tool.
References
1. Pisani, P. H., Lorena, A. C., & De Carvalho, A. C. P. L. F. (2014). Adaptive algo-
rithms in accelerometer biometrics. Proceedings - 2014 Brazilian Conference on In-
telligent Systems, BRACIS 2014, 336–341. http://doi.org/10.1109/BRACIS.2014.67
2. Bifet, Albert, Holmes, Geoff, Kirkby, Richard & Pfahringer, Bernhard (2010). Moa:
Massive online analysis. Journal of Machine Learning Research, 11, 1601-1604.
3. Cruz Silva, N., Mendes-Moreira, J., & Menezes, P. (2013). Features Selection for
Human Activity Recognition with iPhone Inertial Sensors. Advances in Artificial
Intelligence, 16th Portuguese Conference on Artificial Inteligence., 560–570.
4. Zhou, Y., & Goldman, S. (2004). Democratic co-learning. In 16th IEEE International
Conference on Tools with Artificial Intelligence (pp. 594–602). IEEE Comput. Soc.
http://doi.org/10.1109/ICTAI.2004.48
5. Zhou, Z. H., & Li, M. (2005). Tri-training: Exploiting unlabeled data using
three classifiers. IEEE Transactions on Knowledge and Data Engineering, 17(11),
1529–1541. http://doi.org/10.1109/TKDE.2005.186
6. Dawid, A. P. (1984). Statistical theory: the prequential approach. Journal of the
Royal Statistical Society. Series A, 147, 278–292.
7. Kozina, S., Lustrek, M., & Gams, M. (2011). Dynamic signal segmentation for ac-
tivity recognition. Proceedings of International Joint Conference on Artificial Intel-
ligence, 1–12.
8. R Development Core Team (2008). R: A language and environment for statisti-
cal computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN
3-900051-07-0, URL http://www.R-project.org.
9. Tundo, M. D., Lemaire, E., & Baddour, N. (2013). Correcting Smartphone orien-
tation for accelerometer-based analysis. MeMeA 2013 - IEEE International Sym-
posium on Medical Measurements and Applications, Proceedings, (May), 58–62.
http://doi.org/10.1109/MeMeA.2013.6549706