electronics
Review
Identification of Daily Activites and Environments
Based on the AdaBoost Method Using Mobile Device
Data: A Systematic Review
José M. Ferreira 1,† , Ivan Miguel Pires 2,3, *,† , Gonçalo Marques 2,† , Nuno M. Garcia 2,† ,
Eftim Zdravevski 4,† , Petre Lameski 4,† , Francisco Flórez-Revuelta 5,† and
Susanna Spinsante 6,†
1
2
3
4
5
6
*
†
Computer Science Department, Universidade da Beira Interior, 6200-001 Covilhã, Portugal;
jose.ferreira@ubi.pt
Instituto de Telecomunicações, Universidade da Beira Interior, 6200-001 Covilhã, Portugal;
goncalosantosmarques@gmail.com (G.M.); ngarcia@di.ubi.pt (N.M.G.)
Computer Science Department, Polytechnic Institute of Viseu, 3504-510 Viseu, Portugal
Faculty of Computer Science and Engineering, University Ss Cyril and Methodius, 1000 Skopje, Macedonia;
eftim.zdravevski@finki.ukim.mk (E.Z.); petre.lameski@finki.ukim.mk (P.L.)
Department of Computing Technology, University of Alicante, P.O. Box 99, E-03080 Alicante, Spain;
francisco.florez@ua.es
Department of Information Engineering, Università Politecnica delle Marche, 60131 Ancona, Italy;
s.spinsante@staff.univpm.it
Correspondence: impires@it.ubi.pt; Tel.: +351-966-379-785
These authors contributed equally to this work.
Received: 9 December 2019; Accepted: 13 January 2020; Published: 20 January 2020
Abstract: Using the AdaBoost method may increase the accuracy and reliability of a framework for
daily activities and environment recognition. Mobile devices have several types of sensors, including
motion, magnetic, and location sensors, that allow accurate identification of daily activities and
environment. This paper focuses on the review of the studies that use the AdaBoost method with the
sensors available in mobile devices. This research identified the research works written in English
about the recognition of daily activities and environment recognition using the AdaBoost method
with the data obtained from the sensors available in mobile devices that were published between 2012
and 2018. Thus, 13 studies were selected and analysed from 151 identified records in the searched
databases. The results proved the reliability of the method for daily activities and environment
recognition, highlighting the use of several features, including the mean, standard deviation, pitch,
roll, azimuth, and median absolute deviation of the signal of motion sensors, and the mean of the
signal of magnetic sensors. When reported, the analysed studies presented an accuracy higher than
80% in recognition of daily activities and environments with the Adaboost method.
Keywords: daily activities recognition; ensemble learning; ensemble classifiers; environments; mobile
devices; sensors; systematic review
1. Introduction
AdaBoost is one of the first boosting algorithms developed by Yoav Freund and Robert Schapire
that was adapted for practical application in many solving tasks. AdaBoost is a method that uses
ensemble learning techniques to combine multiple weak classifiers into a single strong classifier. It is
combined with other artificial intelligence methods to increase the accuracy of the recognition [1].
Thus, weak learners, including decision tree and decision boosting, are commonly used with the
Electronics 2020, 9, 192; doi:10.3390/electronics9010192
www.mdpi.com/journal/electronics
Electronics 2020, 9, 192
2 of 16
AdaBoost method. In comparison with other machine learning methods, the AdaBoost method is less
susceptible to overfitting.
One of the strategies adopted by the different implementation of Adaboost consists in combination
with other methods to reduce the errors obtained [2,3]. The primary purpose of ensemble
learning techniques is to improve the results by combining the results of different methods [2,3].
These techniques consist of the combination of several machine learning techniques with a single
purpose and model to improve the prediction results [4–6]. It can be divided into two groups,
sequential ensemble methods and parallel ensemble methods, where our focus is the sequential
ensemble methods, because the implementation of Adaboost consists in the application of a base
learner that is generated sequentially [7].
In the last years, several studies have been developed with a focus on the recognition of daily
activities using the sensors available in the commonly used mobile devices. These studies conclude
that it is possible to accurately detect the daily activities and environments with motion, magnetic,
location and acoustic sensors embedded on mobile devices, reporting reliable results available in the
literature with different machine learning methods [8–23].
To date, and due to the increasing power processing capabilities of the different mobile devices,
the Adaboost method is one of the most used methods, and it reports reliable results [24–32].
The motivation of this systematic review is to evaluate the reliability of the Adaboost method for
daily activities and environment recognition using the sensors available in mobile devices for further
implementation of a framework [33–42].
Generally, the raw readings of one-dimensional (e.g., blood pressure sensor, thermometer, etc.) or
multi-dimensional signals (e.g., accelerometer or gyroscope) can be directly processed by AdaBoost,
and other classification and regression algorithms in general. To do that, all sensory readings in a
specific time window represent different inputs. For example, if a thermometer reads data with 1 Hz
frequency, and the window is 60 s, there will be 60 inputs to AdaBoost. Similarly, a three-dimensional
gyroscope would present 180 inputs. Many deep learning methods accept the input data in this
format. Be that as it may. Usually, many algorithms benefit from a feature engineering step [43],
which significantly improves the accuracy or simplifies the complexity of the models [23,44].
Due to the complex nature of the sensory data collected using the sensors available in mobile
devices, the overfitting problem is impacts many machine learning algorithms, including multilayer
perceptron neural networks (MLP), deep neural networks (DNN) and feedforward neural networks
(FNN) [33–42]. Methods for parameter tuning such as grid search [45] and systematic feature
selection [23] are usually applied to mitigate this problem.
Previous studies [33–42] shown that the proposed framework includes the correct modules for
the reliable recognition of daily activities and environments. However, the results can be improved
with other methods, including ensemble learning methods.
This paper reviews the different studies available in the literature related to the implementation
of the AdaBoost method for daily activities recognition. This review is included in the research and
development of a framework associated with the identification of daily activities and environments using
the sensors available in mobile devices, where the AdaBoost method can increase the accuracy compared
to other implementations. The motivation of this paper is to improve the accuracy reported in previous
studies for the recognition. This review intends to explore the use of the Adaboost method to verify if it
reports better results than MLP, FNN, and DNN methods for the identification of daily activities.
The main contribution of this review is the presentation of a base of study for the readers who
deal with the recognition of daily activities and environments using sensors available in mobile devices
providing an in-depth survey of several research projects which implement Adaboost method.
This review shows that the features that reported better results are mean, standard deviation,
pitch, roll, azimuth and median absolute deviation of the signal of motion sensors, and the mean of
the signal of magnetic sensors. According to the results, the Adaboost method provides huge accuracy
for the recognition of daily activities and environments.
Electronics 2020, 9, 192
3 of 16
The following sections are organized as follows: Section 2 presents the methodology of the review.
The results obtained are presented in Section 3. Section 4 presents the discussion on the results. Finally,
the conclusions are presented in Section 5.
2. Methodology
2.1. Research Questions
In this way, the leading questions of this review are: (RQ1) What is AdaBoost? (RQ2) How to
detect daily activities with AdaBoost? (RQ3) How to identify daily activities with AdaBoost using
mobile devices?
2.2. Inclusion Criteria
Studies assessing the recognition of daily living using AdaBoost method were included in this
review according to the following criteria: (1) Detect daily activities using sensors; (2) implementing
AdaBoost method for the automatic recognition of daily activities, presenting the information
about the activities and environments recognized; (3) make use of mobile devices; (4) presents the
accuracies obtained with AdaBoost method; (5) published between 2010 and 2019; (6) were available
in open-access libraries; and (7) written in English.
2.3. Search Strategy
The authors of this review searched for studies according to the inclusion criteria in the following
electronic databases: IEEE Xplore and Science Direct. Every study was independently evaluated by
eight reviewers (JF, IMP, GM, NMG, EZ, PL, FFR, and SS), and all parties evaluated its suitability.
The studies were examined to identify the characteristics of AdaBoost and its relevance for the
implementation in recognition of daily activities and environments using mobile devices.
2.4. Extraction of Study Characteristics
The following data were extracted from the studies and tabulated (see Tables 1 and 2): Year of
publishing, the population was taken into account, purpose, equipment used, and outcomes of each
publication. All cited studies in Tables 1 and 2 informed that the experiments were performed in
laboratory settings. The verification of the availability of the raw data was performed.
Electronics 2020, 9, 192
4 of 16
Table 1. Study summaries.
Authors
Year
Outcomes
Kelarev et al. [46]
2012
A cardiovascular autonomic neuropathy identification algorithm that uses mobile devices is proposed. The dataset has been
created using health records collected in a university research project named Diabetes Complications Screening Research Initiative.
The main contribution of the paper is the recommendation of the AdaBoost and Bagging based on the J48 decision.
Xu et al. [47]
2014
The paper presents an accurate method for context detection, which uses multiple sensors and machine learning. The context
information is restrictively used to select activities that require classification, increasing the accuracy and decreasing the complexity
of the process. Fourteen subjects each carried a tablet, and four 9-DOF sensors were located on wrists, ankle, knee, and mid-waist.
Each volunteer allocated thirty minutes in every context and did each required activity from two to five minutes. The dataset was
then divided into two parts, 30% of the data for training and 70% rest for testing. The combined results of the three classifiers were
able to achieve higher accuracy for all contexts.
Wisniweski et al. [48]
2014
The paper presents an automatic recognition method of asthmatic wheezing through the analysis of a breathing sound dataset.
One hundred thirty s130 records for natural and wheezy breathing using 1024 samples each were used for the study. The overall
recognition was 93%.
Zhou et al. [49]
2015
The authors propose the HATS, which provides both entry-point and post-log-in mobile user authentication. The proposed method
integrates several authentication methods like password, keystroke, gesture, and touch dynamics features to explore the
vulnerabilities of specific approaches to specific security attacks. The participants were required to go through several training
sessions to be introduced to the usage of two different keyboards. Twelve volunteers (for men, eight women) carried the study.
Masri et al. [50]
2015
The study proposes active authentication applying scrolling behaviors for biometrics and evaluates diverse classification and clustering
approaches that support those characteristics. The experiment counted with 84 participants and 54 documents. The most accurate
method was achieved adopting k-means clustering among two techniques applied to validate users, with a success rate of 83.5%.
Xu et al. [51]
2016
The authors propose an online learning approach for activity recognition based on data collected using inertial sensors. The data
was gathered from fourteen volunteers. Every volunteer performs thirty minutes in the respective context and carried each
required activity for two to five minutes. This algorithm outperformed the benchmark algorithms by 30–40%.
Tang et al. [52]
2016
The paper shows an assessment of ten representative classifiers applied in two datasets. The dataset contains accelerometer
time-series data from 22 volunteers. This study concluded that K-Nearest Neighbors is the most suitable classifier.
Yanyun et al. [53]
2017
The paper presents a method based on Convolutional Neural Networks approach to provide automatic extraction of features for
transportation mode classification. There were used a total of 169 features, and the dataset has more than 200 h of transportation data
collected from thirty volunteers on diverse transportation modes (bus, car, metro, train). The recognition accuracy was: 96.6% for the
bus, 99.6% for the car, 99.0% for the metro, and 98.9% for the train. Giving an average accuracy of 98.6%.
Electronics 2020, 9, 192
5 of 16
Table 1. Cont.
Authors
Year
Outcomes
Li et al. [54]
2017
The authors propose an indoor and outdoor recognition method, which is divided into two parts: The machine learning-based
Indoor, Outdoor, and Semi-open areas recognition algorithm and the lightweight WiFi sub-detector. The absolute values and the
relative measurements of WiFi received signal strength are calculated to identify if the user environment is a semi-open area,
indoor or outdoor. The proposed method presents 85% of accuracy for the lightweight WiFi-based technique and 96% of accuracy
using the aggregated IOS-detector.
Yanjun et al. [55]
2017
The article proposes a Bayesian algorithm for traffic pattern recognition. The used dataset consists of 400 h from eight individuals. An
accelerometer, a barometer, a geomagnetic, a gyroscope, and base station were the five used sensors. The AdaBoost classification
method was also implemented to get better results. The proposed method presents an accuracy rate from 83.3% to 91.5%.
Vafeiadis et al. [56]
2017
The paper presents a machine learning approach for occupancy detection. The water and energy consumption data collected using
smart meters are used as features for occupancy detection in a domestic environment. Under their boosting versions, Random
Forest and Decision Tree classifiers present more accuracy when associated with the other classifiers. The authors obtain an overall
accuracy of 83.37% and 82.79%, respectively.
Subasi et al. [57]
2018
This study proposes the use of AdaBoost based classifier for human activity recognition using data collected from sensors located
on the body. The study is based on nine inertial sensors collected by seventeen volunteers who perform 33 fitness exercises.
The results present 99.98% of success rate.
Yuan et al. [58]
2018
The authors present an indoor localization algorithm based on ’Twi-AdaBoost’. The proposed method uses several sensors, such as
gyroscope, magnetometer, and accelerometer. The tests used 6304 samples collected from both smartphone and smartwatch
devices. The AdaBoost method outperforms the other approaches tested in every metrics.
Table 2. Critical analysis of reviewed studies.
Authors
Population
Purpose of Study
Device Type
Public Dataset
Pros
Cons
Kelarev et al. [46]
Not Mentioned
Cardiovascular autonomic neuropathy
identification on mobile devices
Mobile Devices
Yes
Evaluated multiple ensemble strategies;
Novel ensemble of AdaBoost and Bagging
based on J48
Dataset is scarcely described;
Processing pipeline not illustrated;
Not evaluated on different test subjects
Xu et al. [47]
Fourteen individuals use an
Android tablet, and four 9-DOF
(Degrees of freedom) sensors were
located on the wrist, knee, ankle,
and mid-waist
An end-to-end system is proposed to
enable large-scale supervision and
classification of physical movements
Mobile Devices
Yes
End-to-end system that integrates context
into activity classification; 13 activities of
daily living; 8 environments;
Battery life considerations
The authors should use models able to
learn the activities associated with
contexts in conjunction with scenarios;
The authors should test the robustness
of the system and improve privacy,
security, and user friendliness;
Not evaluated on different test subjects
Electronics 2020, 9, 192
6 of 16
Table 2. Cont.
Authors
Population
Purpose of Study
Device Type
Public Dataset
Pros
Cons
Wisniweski et al. [48]
Population not mentioned,
130 records for natural and wheezy
breathing using 1024 samples each
An automatic and highly efficient method
is proposed for asthmatic wheezing
recognition in breathing sounds
Mobile Devices
Yes
It detects asthmatic wheezes;
The implemented method is not
computationally complex; it is capable of
monitoring asthma using mobile devices
The authors should test other machine
learning methods; Not evaluated on
different test subjects; Dataset is
scarcely described
Zhou et al. [49]
Twelve individuals (four men and
eight women)
A harmonized user authentication method
based on ThumbStrokes dynamics (HATS)
for smartphone touch screen
Smartphones
No
It supports entry-point and post-login mobile
user authentication; It explores previous
studies to implement a better solution;
It improves the security and authentication of
mobile authentication systems
The study needs to be tested using a larger
sample size; The authors does not perform
feature reduction, feature selection and
data transformation; Not evaluated with
different test subjects and devices
Masri et al. [50]
84 individuals participated in the
experiment and 54 documents were
available for reading
Discusses active authentication methods
which utilise scrolling behaviors for
bio-metrics and evaluates diverse
classification, and clustering approaches
that lead to those characteristics
Mobile Devices
Yes
The data used is composed by data
acquired from the different events related
to the users’ reading habits;
Novel non-intrusive approach for active
and continuous re-authentication; It took
into account the security of the mobile
application; The model is capable of
accessing control, intrusion detection and
recommender systems; The models test the
authentication after x scrolls
The model should be the model
changed to test for authentication after
x amount of scrolls; Not evaluated on
different test subjects; Dataset is
scarcely described
Xu et al. [51]
Fourteen individuals carried in the
research test. Every individual uses
thirty minutes in every context.
Moreover, the participants spend
two to five minutes in each required
activity under every context
An activity recognition method using
contextual online learning techniques
using data collected by low-cost inertial
sensors is presented
Smartphones
No
The proposed algorithm outperforms the
existing algorithms without requiring
training phase from the individual;
The algorithm is rigorously characterized;
The method is capable of performing the
recognition of activities with an online,
personalized and adaptive method
The use of context did not significantly
improve the results; Not evaluated with
different test subjects and devices
Tang et al. [52]
22 individuals
The authors have tested ten classifiers
on two public datasets for user
activity recognition.
Mobile Devices
No
The authors used two published datasets
with different activities; The datasets are
clearly described; The authors tested
different artificial intelligence methods to
identify walking patterns
The authors should extend the dataset
with other sensors; Not evaluated with
different test subjects and devices
Yanyun et al. [53]
Thirty individuals. The test has
used a total of 169 features.
75 horizontal and 22 vertical
acceleration features; 22 triaxial
acceleration magnitude features;
22 gyroscope and 22 geomagnetic
features; and 6 atmospheric
pressure features
A transportation mode recognition method
based on Convolutional Neural Networks
is presented
Smartphones
Yes
Convolutional Neural Networks used
automatically learns the different
transportation modes; It was performed the
shallow feature extraction before the
learning deep feature; The implemented
method reduces the complexity; It is useful
for mobile devices; A small number of layer
works well for the pre-processing of the data
The authors should consider
context-aware technologies;
Not evaluated with different test
subjects and devices
Li et al. [54]
Not mentioned
Proposes a lightweight indoor, outdoor and
Semi-open recognition algorithm
Smartphones
No
It may realize localization indoor to
outdoor, and vice versa; It presents the
workflow of the algorithm implemented;
The authors customized some methods to
improve the energy efficiency
The authors should incorporate
run-time verification methods to
improve the accuracy and safety of the
proposed method; Not evaluated with
different test subjects and devices
Electronics 2020, 9, 192
7 of 16
Table 2. Cont.
Authors
Population
Purpose of Study
Device Type
Public Dataset
Pros
Cons
Yanjun et al. [55]
Eight individuals
Proposes a traffic pattern recognition
method based on the Bayesian algorithm to
identify the traffic
Smartphones
Yes
The algorithm used the sensors available in
the mobile devices to detect traffic patterns;
400 h of data; The data was collected from
different cities in the world;
The transportation modes are presented
The proposed method presents very
low accuracy when using only the
acceleration sensor data. In this case,
the distinction between the car and bus
cannot performed; Not evaluated with
different test subjects and devices
Vafeiadis et al. [56]
Three individuals
Proposes an occupancy detection
algorithms using machine learning.
The dataset consists of water and energy
consumption information collected from
smart meters
Mobile Devices
Yes
The use of smart meters for different
sensing; The use of feature selection to
discover the most important features;
The authors used real-time data; The
authors implemented several machine
learning methods; It was used the boosting
technique; The activities/environments are
identified; Three individuals acquired data
during one month; Ensemble classifiers
achieved better performed than others,
because it combines different classifiers;
Feature selection helped to reduce the
dataset sparsity
The authors must improve the dataset
and the features used as input to the
classifiers. Feature selection methods
must be incorporated to enhance the
accuracy of the predictive model.
Subasi et al. [57]
Seventeen participants have
performed 33 fitness activities and
use nine inertial sensors
Proposes an activity recognition method
which uses data collected from sensors
located on the body
Smartphones
Yes
The authors used the 10-fold
cross-validation to test the algorithm;
33 fitness activities; The methods and
workflow are clearly described
The study uses a high number of
sensors; the authors must implement an
attribute selection method due to the
use of numerous sensors resulting in
117 attributes for each instance;
Not evaluated with different test
subjects and devices
Yuan et al. [58]
Not mentioned. The dataset consists
of more than 36000 samples
Proposes an indoor localization method
using mobile sensors
Mobile Devices
Yes
The authors implemented different
combinations of Adaboost method;
36000 samples collected in a real-world
environment; The features are clearly
described; The proposed method
outperforms the state-of-the-art,
presenting low errors
The correlation between position x and
y in the same location is not performed
by the authors to improve accuracy;
Not evaluated with different test
subjects and devices
Electronics 2020, 9, 192
8 of 16
3. Results
As pictured in Figure 1, we identified 151 papers with three duplicates, that were removed.
The other 148 articles were evaluated according to the title, keywords, and abstract, excluding
133 citations. After full-text evaluation, two papers were removed from the remaining 15 papers.
The qualitative and quantitative synthesis included information related to the remaining 13 articles.
In conclusion, we examined 13 documents.
Figure 1. Articles analysis.
To find relevant information about the implementations presented in the different studies analysed
in this review, the reader should find the information in the original cited works. Table 1 shows the year
of publication and the resume of the papers and final results. Table 2 shows the population, the purpose
of the study, devices, settings of the papers, pros, and cons. When the datasets used in a study is publicly
available, or the population information is provided, it is considered as a positive aspect. In many cases,
the evaluation uses a cross-validation scheme (regular or stratified per class). However, the studies do not
consider different subsets of the population for training and testing (i.e., train/test split based on subjects
or patients). This is generally a more rigorous evaluation scheme and is expected to hurt the reported
accuracy. Other more specific pros and cons are provided for each study.
The papers were published between 2012 and 2018, where two studies were published in 2018
(15%), four studies were published in 2017 (31%), two studies were published in 2016 (15%), two studies
were published in 2015 (15%), two studies were published in 2014 (15%), and one study was published in
2012 (8%). Regarding the used devices, it was split among 43% for smartphones and the remaining 57%
Electronics 2020, 9, 192
9 of 16
for mobile devices. The source code is not available for all studies analysed. Moreover, 69% of the studies
have the raw data available. Finally, we verified that there are no studies that shared the source code.
Methods for Identification of Activities in Daily Living
In the study [57], the authors tried to use different classifiers for the recognition of activities with
sensors to find the best method. Ten classifiers were utilized with the AdaBoost method. The dataset
used was publicly available. The settings were investigated using nine inertial sensors from seventeen
individuals taking into account 33 fitness activities. The used sampling rate was 50 Hz. After checking
accuracies of the AdaBoost method, authors came to conclude that its implementation with random
forest gives the best accuracy, with a value of 99.98%.
Authors of [49] have proposed harmonized authentication based on ThumbStroke dynamics
(HATS) for mobile devices. The performance of HATS was tested, taking into account the different
screen sizes of several mobile devices. Laboratory experiments were conducted to collect data for
testing. Participants were required prior experience with touch screen devices and a qwerty keyboard.
The study selected some features for learning ThumbStroke models, and these are timing features,
spatial features, movement direction features, and operation features. The phrases, entered by the
participants, were adopted from MacKenzie and Soukoreff and varied from 16 to 43 characters.
Based method across all settings and classification models, the final results showed that HATS
outperformed the keystroke dynamics. Among all the classification methods used, AdaBoost reported
a maximum accuracy of 41.8%.
Li et al. [54] talks about an indoor/outdoor detection system (IOS). This method is split by the
machine learning-based IOS-detector and the lightweight WiFi sub-detector. The first part infers indoor,
outdoor, or semi-open environments based on the classification results. The second part focuses on
the implementation of mobile devices. Finally, the other part consists of the IOS detection that shows
high accuracy for the system. In conclusion, the proposed IOS detector achieves around 96% for the
aggregated IOS detector and over 85% accuracy for the lightweight WiFi-based sub-detector.
In the study [50], the authors introduce a method for re-authenticating users taking into account
a behavioral biometric-based on users’ document scrolling traits. More specifically focused on
identifying abnormal scrolling behavior on users while interacting with protected or read-only
documents. Dataset was obtained from a previous project aimed to detect document access activities
that indicate cyber attacks. Features for this paper were slit in vectors, being vector one derived from
scrolling traits, vector two a representation of the polarity of scrolling, and vector 3 treats the dataset
as a bipartite graph with two node sets. k-means clustering achieved the best performance with an
83.5% success rate in predicting the authenticated user.
The paper [48] presents a highly efficient method for the automatic detection of asthmatic
wheezing in breathing sounds. The process is suitable for personal asthma monitoring via mobile
devices since its not computationally complex. Most of the used data came from online databases of
Human lung sounds. However, the authors also used several of their recordings of regular and wheezy
breaths. The authors also confirmed the optimality of the audio spectral envelope (ASE) plus the
value of the tonality index (TI) as a feature detector, using the mRMR (minimal redundancy–maximal
relevance) method. Thousands of experiments were performed, and the best results were obtained
from the fluctuation of the Audio Spectral Envelope descriptor adopted from the MPEG-7 standard,
reporting an accuracy around 100%.
Authors of [53] developed a method to collect the sensor data, acceleration, gyroscope,
geomagnetic, and atmospheric pressure were the four kinds of sensors used. The shallow feature
extraction of the raw data happens before the CNN learning deep feature, which will reduce the
complexity of the network and training time of the model. This process is critical for smartphones
because of their limited resources. Three classes of features are extracted from each frame,
including statistical, time, and frequency domains. Namely, the features used are: Mean, standard
deviation, variance, median, minimum, maximum, range, interquartile range, kurtosis, skewness, root
Electronics 2020, 9, 192
10 of 16
mean square, integral, double integral, autocorrelation, mean-crossing rate, fast Fourier transform,
spectral energy, spectral entropy, spectrum peak position, wavelet entropy, and wavelet magnitude.
Final results show that the proposed method can achieve 98% accuracy, meaning it outperforms
the SVM (support vector machine) and AdaBoost classification in efficiency and computational cost,
reporting accuracy of 93.6% with AdaBoost.
Yuan et al. [58] propose an indoor localization system using sensors for smartphones and
smartwatches. Over 36,000 samples of data were collected in a 185.12 m2 real indoor environment by
a user using two different devices. Looking with the experimental results, the authors concluded that
Twi-AdaBoost outperforms the state-of-the-art indoor localization algorithms. The localization error
of position x and y achieved was 0.387 m and 0.398 m, respectively. The used datasets include the
features: Place ID, Timestamp, Accelerometer_X, Accelerometer_Y, Accelerometer_Z, MagneticField_X,
MagneticField_Y, MagneticField_Z, X_Axis Angle (Pitch), Y_Axis Angle (Roll), Z_Axis, Angle (Azimuth),
Gyroscope_X, Gyroscope_Y, and Gyroscope_Z, reporting an accuracy around 99%.
In the paper [55], a novel technique based on the Bayesian voting algorithm that can be used with
low-power sensors for transportation mode detection is presented. The authors used a set of data that
consists of 400 h from eight individuals. Five sensors were used, being those: Acceleration, gyroscope,
geomagnetic, barometer, and base station obtain by using AdaBoost classification to improve the results.
Besides, the Bias algorithm was used to extract the features to reduce the adaptive boosting feature
dimensions and determine the critical factors for identifying different transportation modes. The features
used are: Mean, standard deviation, variance, median, minimum, maximum, range, interquartile,
kurtosis, skewness, root mean square, time integral, double integral, auto-correlation, mean-crossing rate,
fast Fourier transform, spectral energy, spectral entropy, spectrum peak position, wavelet entropy, wavelet
magnitude, peak volume, intensity, length, variance of peak features, peak frequency, stationary duration,
stationary frequency. Taking into account the final results, authors concluded that their algorithm could
supply and replace some traffic pattern recognition algorithms and fix the problem that different mobile
phones have various sensors, reporting accuracy between 64.54% and 96.83%.
In [51], the authors presented a contextual multi-armed bandits (MAB) approach that enables
activity classification. This method makes context adaptation, continuous online learning, and active
learning. Since the cost of extracting specific features is very high, the authors decided to use side
information as the context. Since features can be used as contexts, this is not a limitation for the project.
The proposed algorithm with active learning outperformed the benchmark algorithms by an average
of 35%, reporting, and accuracy between 70% and 85%.
Xu et al. [47] focuses on three challenges, including the ability to accurately detect context using
sensors and machine learning. The selection of activities for classification is performed by using
context, reducing the complexity and improving the accuracy, speed, and energy usage, and the
ability for experts in prescribing sets of physical activities under different environments. The features
used for the project were: kNN (k-Nearest Neighbor) with time, kNN with wireless media access
control (MAC) address and signal strength, and AdaBoost with audio peak frequency, peak energy,
average power, and total energy. These were extracted from raw sensor data using a java program
implementing the IContextFeatureExtractor interface. The data used was acquired by 14 participants
that carried an Android mobile phone, and four 9-DOF devices were placed on dominant wrists, knee,
ankle, and mid-waist. Each subject performed every required activity under every context for 2–5 min.
The data were split into training (30%) and testing (70%) sets. Authors concluded that despite the
methodology demonstrating effectiveness, efficiency, and potential, a more extensive study needs to
be performed to improve privacy, security, and user-friendliness, reporting accuracy between 59% and
100%.
In [56], the problem of occupancy detection in a domestic environment was studied using machine
learning techniques and their boosting versions on a dataset collected from electricity and water
consumption smart meters. These features were selected using the Mutual Information technique.
The dataset contains energy and water consumption (during summer) time data of 1-minute resolution
Electronics 2020, 9, 192
11 of 16
for 16 consecutive days. The features included in the used dataset were: Central power, refrigerator,
television, washing machine, dryer, cold water-kitchen, hot water-kitchen, dishwasher-water and
washing machine-water, reporting accuracy higher than 70%.
Authors of [52] evaluated ten representative classifiers in the identification of two available
datasets. The first dataset consists of accelerometer readings of walking patterns from 22 participants.
The second one contains activity and postural transition data collected from the accelerometer and
magnetometer data acquired from 30 participants. For the Walking dataset, the authors split the data
into fixed-width sliding windows with a 50% overlap and extract nine features from every window and
scale the features to [−1, 1]. The authors obtained the mean, standard deviation, and median absolute
deviation from the different axis of the sensors. The authors of the study already pre-processed the
sensor signals by noise filter and partitioned the data into fixed-width sliding windows with a 50%
overlap as well and constructed a 561-feature vector for every window. From those features, authors
extracted 24 features, including mean, standard deviation from the different axis of body acceleration,
gravity acceleration, jerk signals of body acceleration, angular velocity, and jerk signals of angular
velocity. In conclusion, the authors reported an accuracy between 95.6% and 97.8%.
The study [46] focuses on using mobile devices for the detection of cardiovascular autonomic
neuropathy. The authors concentrated on the task of the detection and monitoring of cardiovascular
autonomic neuropathy. After all the studies, they concluded that best outcomes were obtained by
the novel combined ensemble of AdaBoost and Bagging based on the J48 decision tree, reporting the
highest accuracy of 94.53%.
4. Discussion
This review confirms that AdaBoost, and in general boosting ensemble methods, are reliable for
the identification of daily activities. Several studies are not well described, and the source code of the
algorithms are not publically available. The verification and reproducibility of the obtained results
is not easily possible, because of the following reasons: Only some authors shared the datasets; in
many cases, the methods are not explained well explained, in particular, the preprocessing of the
datasets; and the hyper-parameter tuning is poorly described, or the exact algorithm parameters are
not described.
The number of studies using the AdaBoost method for the recognition of daily activities is
minimal, and the daily activities mainly recognized are the simple activities, including walking,
running, walking upstairs and downstairs, and other quotidian activities.
Following our literature review, most of the analysed studies (85%) report the best results using
AdaBoost methods. Only two studies (15%) presented in [49,58] have said that the AdaBoost based
methods do not show the best results when compared with the other approaches for daily activities
and environments recognition. Nevertheless, the authors of these studies still recognised the reliable
applicability of the AdaBoost method for activity and environment recognition activities.
In summary, all reviewed works first perform a feature extraction step, which somewhat varies
depending on the used sensor types. In cases of multiple sensors, or multi-channel sensors, the feature
extraction is performed independently for each time series (i.e., channel or sensor). Generally, various
statistical metrics, as listed in Table 3, are computed on the raw signal in the time domain, and rarely
features are deriving from the frequency domain. Then, after the features are extracted from each sensor
as a separate time series, the extracted features are fed into the classifiers. Very often, a systematic
approach to feature extraction improves the accuracy [23].
The authors used different features, and the average accuracies obtained with them can be
comparable. Table 3 presents the average accuracy of the various features extracted, verifying that the
features that allow the recognition of daily activities with an accuracy higher than 90% are the mean,
standard deviation, pitch, roll, azimuth and median absolute deviation of signal of motion sensors,
and the mean of the signal of magnetic sensors.
Electronics 2020, 9, 192
12 of 16
Table 3. Average of the accuracy reported in the studies analysed, grouped by features.
Feature
Average reported accuracy
with AdaBoost
mean of signal of magnetic sensors
99.0%
pitch, roll, and the azimuth of the signal of motion sensors
99.0%
median absolute deviation of the signal of motion sensors
96.7%
mean of signal of motion sensors
96.0%
standard deviation of the signal of motion sensors
90.1%
median, variance, minimum and maximum values, interquartile range,
range, skewness, kurtosis, integral, double-integral, Root Mean Square
(RMS), Fast Fourier Transform (FFT), spectral entropy, spectral energy,
wavelet entropy, spectrum peak position and wavelet magnitude of
signal of motion sensors
87.1%
scrolling traits and polarity of scrolling
83.5%
peak volume, intensity, variance of peaks, stationary duration and
stationary frequency of the signal of motion sensors
80.6%
peak frequency of the signal of motion sensors
80.3%
peak energy, average power and total energy of signal of motion sensors
80.0%
Moreover, Table 4 presents the advantages and disadvantages of the Adaboost method, proving
that it can be used for the recognition of daily activities and environments with the recent advancements
in the hardware and software of the devices commonly used.
Table 4. Advantages and disadvantages of the use of Adaboost method in the different studies analyzed.
Pros
-
Cons
The combination of the Adaboost and J48 decision
tree revealed the best results.
Adaboost can be used for the monitoring of diabetes.
Adaboost with Bagging and Boosting based on
decision trees reported reliable accuracy.
This algorithm can be applied for real-time
assessments with sensor data.
It provides high recognition accuracy and low
computational complexity.
It provides high security and usability of the
different implementations.
It can be executed in real-time with reliable accuracy.
The combination of Adaboost with the k-Nearest
Neighbors algorithm outperformed all other classifiers.
The Adaboost method shows high reliability in the
recognition of different activities.
The results obtained can be correlated between
different devices.
-
The research on multi-level classifiers should
continue to improve the results.
The energy consumption of the Adaboost
method is very high.
It should always have high reliability for
medical purposes.
It has limited capabilities for recognition.
The classified should be updated with new data.
Larger-scale experiments need to be conducted
to validate the efficacy of the algorithms further.
In comparison with other algorithms, the Adaboost method uses different algorithms as the weak
learner, in which these algorithms will take into account the features extracted from the signals, such as
mean, standard deviation, variance, and others. In general, Adaboost made use of complex data,
but it can be used with 1D data in comparison with other algorithms. The authors of the research
studies analysed used the Adaboost with uni-dimensional data, i.e., they used the features extracted
from the data to provide the results, where the results obtained proved its reliability for physical and
physiological data.
In conclusion, the use of mobile devices for daily activities recognition using AdaBoost is limited,
because of the low power processing and battery capabilities of these devices [59,60]. According to
Electronics 2020, 9, 192
13 of 16
the reported studies in this review, it is possible to conclude that the use of the AdaBoost method is
reliable with mobile devices as verified by the accuracies reported in the different studies, where only
two studies reported accuracies lower than 50%.
5. Conclusions
This review presents studies available in the literature that use the AdaBoost method for the
recognition of daily activities and environments. Thirteen studies were analysed, and the main findings
are summarised as follows:
•
•
•
(RQ1) The AbaBoost method is an ensemble learning method that is used in conjunction with
other algorithms. The different algorithms are commonly named as weak classifiers, avoiding the
overfitting problem;
(RQ2) The AdaBoost method is implemented in conjunction with other algorithms to increase the
accuracy of the recognition of daily activities and environments;
(RQ3) For the recognition of daily activities and environments, the AdaBoost method is combined
with a weak classifier. The features that reported better accuracy are the mean, standard deviation,
pitch, roll, azimuth, and median absolute deviation of the signal of motion sensors, and the mean
of the signal of magnetic sensors.
This review also highlights the use of smartphones and other mobile devices as they should have
a particular purpose because of limited battery life and processing capabilities. First, the authors
excluded studies that are not focused on the recognition of daily activities end environments with
the AdaBoost method. Secondly, the studies that do not use sensors available on mobile devices
were excluded. We excluded several studies after analysis of the abstracts and full-text of the papers.
Another reason for exclusion was the language of the study, excluding the studies that were not written
in English. With the features collected, the AdaBoost method allows recognition with an accuracy
higher than 80%.
As future work, the implementation of the AdaBoost method in the framework for the recognition
of daily activities and environments; it will be used to recognize seven daily activities and nine
environments.
Conceptualization, methodology, software, validation, formal analysis,
Author Contributions:
investigation,writing—original draft preparation, writing—review and editing: J.M.F., I.M.P., G.M., N.M.G., E.Z.,
P.L., F.F.-R. and S.S. All authors have read and agreed to the published version of the manuscript.
Funding: This work is funded by FCT/MEC through national funds and when applicable co-funded by
FEDER—PT2020 partnership agreement under the project UIDB/EEA/50008/2020 (Este trabalho é financiado pela
FCT/MEC através de fundos nacionais e quando aplicável cofinanciado pelo FEDER, no âmbito do Acordo de Parceria
PT2020 no âmbito do projeto UIDB/EEA/50008/2020).
Acknowledgments: This work is funded by FCT/MEC through national funds and when applicable co-funded
by FEDER—PT2020 partnership agreement under the project UIDB/EEA/50008/2020 (Este trabalho é financiado
pela FCT/MEC através de fundos nacionais e quando aplicável cofinanciado pelo FEDER, no âmbito do Acordo de
Parceria PT2020 no âmbito do projeto UIDB/EEA/50008/2020). This article is based upon work from COST Action
IC1303-AAPELE-Architectures, Algorithms and Protocols for Enhanced Living Environments and COST Action
CA16226-SHELD-ON-Indoor living space improvement: Smart Habitat for the Elderly, supported by COST
(European Cooperation in Science and Technology). More information in www.cost.eu.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
2.
3.
Schapire R.E. Explaining AdaBoost. In Empirical Inference; Schölkopf, B., Luo, Z., Vovk, V., Eds.; Springer:
Berlin/Heidelberg, Germany, 2013.
Webb, G.I.; Zheng, Z. Multistrategy ensemble learning: Reducing error by combining ensemble learning
techniques. IEEE Trans. Knowl. Data Eng. 2004, 16, 980–991. [CrossRef]
Lorena, A.C.; De Carvalho, A.C.; Gama, J.M. A review on the combination of binary classifiers in multiclass
problems. Artif. Intell. Rev. 2008, 30, 19. [CrossRef]
Electronics 2020, 9, 192
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
14 of 16
Ganjisaffar, Y.; Caruana, R.; Lopes, C.V. Bagging gradient-boosted trees for high precision, low variance
ranking models. In Proceedings of the 34th International ACM SIGIR Conference on Research and
Development in Information Retrieval, Beijing, China, 25–29 July 2011; pp. 85–94.
Lee, B.K.; Lessler, J.; Stuart, E.A. Improving propensity score weighting using machine learning. Stat. Med.
2010, 29, 337–346. [CrossRef] [PubMed]
Yang, P.; Hwa Yang, Y.; B Zhou, B.; Y Zomaya, A. A review of ensemble methods in bioinformatics.
Curr. Bioinform. 2010, 5, 296–308. [CrossRef]
Zhou, Z.-H. Ensemble Methods: Foundations and Algorithms; CRC Press: Boca Raton, FL, USA, 2012.
Akhoundi, M.A.A.; Valavi, E. Multi-sensor fuzzy data fusion using sensors with different characteristics.
arXiv 2010, arXiv:1010.6096.
Banos, O.; Damas, M.; Pomares, H.; Rojas, I. On the use of sensor fusion to reduce the impact of rotational
and additive noise in human activity recognition. Sensors 2012, 12, 8039–8054. [CrossRef] [PubMed]
Dernbach, S.; Das, B.; Krishnan, N.C.; Thomas, B.L.; Cook, D.J. Simple and complex activity recognition
through smartphones. In Proceedings of the 2012 Eighth International Conference on Intelligent
Environments, Guanajuato, Mexico, 26–29 June 2012; pp. 214–221.
Hsu, Y.; Chen, K.; Yang, J.; Jaw, F. Smartphone-based fall detection algorithm using feature extraction.
In Proceedings of the 2016 9th International Congress on Image and Signal Processing, BioMedical
Engineering and Informatics (CISP-BMEI), Datong, China, 15–17 October 2016; pp. 1535–1540.
Paul, P.; George, T. An effective approach for human activity recognition on smartphone. In Proceedings of the
IEEE International Conference on Engineering and Technology (ICETECH), Coimbatore, India, 20 March 2015.
Shen, C.; Chen, Y.; Yang, G. On motion-sensor behavior analysis for human-activity recognition via
smartphones. In Proceedings of the 2016 IEEE International Conference on Identity, Security and Behavior
Analysis (ISBA), Sendai, Japan, 29 Feburary–2 March 2016.
Wang, J.; Chen, Y.; Hao, S.; Peng, X.; Hu, L. Deep learning for sensor-based activity recognition: A survey.
Pattern Recognit. Lett. 2019, 119, 3–11. [CrossRef]
Gomaa, W.; Elbasiony, R. Comparative Study of Different Approaches for Modeling and Analysis of
Activities of Daily Living. SSRN Electron. J. 2019. [CrossRef]
Qi, J.; Yang, P.; Newcombe, L.; Peng, X.; Yang, Y.; Zhao, Z. An overview of data fusion techniques for Internet
of Things enabled physical activity recognition and measure. Inf. Fusion 2019, 55, 269–280. [CrossRef]
Garcia, N.M.; Rodrigues, J.J.P. (Eds.) Ambient Assisted Living; CRC Press: Boca Raton, FL, USA, 2015.
Garcia, N.M. Roadmap to the Design of a Personal Digital Life Coach. In International Conference on ICT
Innovations; Springer: Cham, Switzerland, 2015; pp. 21–27.
Sousa, P.S.; Sabugueiro, D.; Felizardo, V.; Couto, R.; Pires, I.; Garcia, N.M. mHealth sensors and applications
for personal aid. In Mobile Health; Springer: Cham, Switzerland, 2015; pp. 265–281.
Dobre, C.; Mavromoustakis, C.X.; Garcia, N.M.; Mastorakis, G.; Goleva, R.I. Introduction to the AAL and ELE
Systems. In Ambient Assisted Living and Enhanced Living Environments; Butterworth-Heinemann: Oxford, UK, 2017.
Felizardo, V.; Sousa, P.; Sabugueiro, D.; Alexandre, C.; Couto, R.; Garcia, N.; Pires, I. E-Health: Current status
and future trends. In Handbook of Research on Democratic Strategies and Citizen-Centered E-Government Services;
IGI Global: Hershey, PA, USA, 2015; pp. 302–326.
Goleva, R.I.; Garcia, N.M.; Mavromoustakis, C.X.; Dobre, C.; Mastorakis, G.; Stainov, R.; Chorbev, I.;
Trajkovik, V. AAL and ELE Platform Architecture. In Ambient Assisted Living and Enhanced Living
Environments; Butterworth-Heinemann: Oxford, UK, 2017; pp. 171–209.
Zdravevski, E.; Lameski, P.; Trajkovik, V.; Kulakov, A.; Chorbev, I.; Goleva, R.; Pombo, N.; Garcia, N.
Improving Activity Recognition Accuracy in Ambient-Assisted Living Systems by Automated Feature
Engineering. IEEE Access 2017, 5, 5262–5280. [CrossRef]
Aguileta, A.A.; Brena, R.F.; Mayora, O.; Molino-Minero-Re, E.; Trejo, L.A. Multi-Sensor Fusion for Activity
Recognition—A Survey. Sensors 2019, 19, 3808. [CrossRef] [PubMed]
Esmaeili Kelishomi, A.; Garmabaki, A.H.S.; Bahaghighat, M.; Dong, J. Mobile User Indoor-Outdoor Detection
through Physical Daily Activities. Sensors 2019, 19, 511. [CrossRef] [PubMed]
Deep, S.; Zheng, X.; Karmakar, C.; Yu, D.; Hamey, L.; Jin, J. A Survey on Anomalous Behavior Detection for
Elderly Care using Dense-sensing Networks. IEEE Commun. Surv. Tutor. 2019. [CrossRef]
Qolomany, B.; Al-Fuqaha, A.; Gupta, A.; Benhaddou, D.; Alwajidi, S.; Qadir, J.; Fong, A.C. Machine Learning,
Big Data, And Smart Buildings: A Comprehensive Survey. arXiv 2019, arXiv:1904.01460.
Electronics 2020, 9, 192
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
15 of 16
Nweke, H.F.; Teh, Y.W.; Mujtaba, G.; Al-Garadi, M.A. Data fusion and multiple classifier systems for human
activity detection and health monitoring: Review and open research directions. Inf. Fusion 2019, 46, 147–170.
[CrossRef]
Moustaka, V.; Vakali, A.; Anthopoulos, L.G. A systematic review for smart city data analytics. ACM Comput.
Surv. (CSUR) 2019, 51, 103. [CrossRef]
Nweke, H.F.; Teh, Y.W.; Mujtaba, G.; Alo, U.R.; Al-garadi, M.A. Multi-sensor fusion based on multiple
classifier systems for human activity identification. Hum.-Cent. Comput. Inf. Sci. 2019, 9, 34. [CrossRef]
Eakin, P. Problems with assessments of activities of daily living. Br. J. Occup. Ther. 1989, 52, 50–54. [CrossRef]
Law, M. Evaluating activities of daily living: Directions for the future. Am. J. Occup. Ther. 1993, 47, 233–237.
[CrossRef]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F. From Data Acquisition to Data Fusion:
A Comprehensive Review and a Roadmap for the Identification of Activities of Daily Living Using Mobile
Devices. Sensors 2016, 16, 186. [CrossRef]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F. Identification of activities of daily living using
sensors available in off-the-shelf mobile devices: Research and hypothesis. In Proceedings of the International
Symposium on Ambient Intelligence, Seville, Spain, 1–3 June 2016; pp. 121–130.
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F.; Spinsante, S. Pattern recognition techniques for the
identification of Activities of Daily Living using mobile device accelerometer. arXiv 2017, arXiv:1711.00096.
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F.; Spinsante, S.; Goleva, R.; Zdravevski, E. Recognition
of activities of daily living based on environmental analyses using audio fingerprinting techniques:
A systematic review. Sensors 2018, 18, 160. [CrossRef] [PubMed]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F.; Spinsante, S. Approach for the development of a
framework for the identification of activities of daily living using sensors in mobile devices. Sensors 2018, 18, 640.
[CrossRef] [PubMed]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F.; Spinsante, S.; Teixeira, M.C. Identification of
activities of daily living through data fusion on motion and magnetic sensors embedded on mobile devices.
Pervasive Mob. Comput. 2018, 47, 78–93. [CrossRef]
Pires, I.M.; Teixeira, M.C.; Pombo, N.; Garcia, N.M.; Flórez-Revuelta, F.; Spinsante, S.; Goleva, R.; Zdravevski,
E. Android Library for Recognition of Activities of Daily Living: Implementation Considerations, Challenges,
and Solutions. Open Bioinform. J. 2018, 11, 61–88. [CrossRef]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F. Framework for the Recognition of Activities of
Daily Living and their Environments in the Development of a Personal Digital Life Coach. In Proceedings
of the DATA 2018: 7th International Conference on Data Science, Technology and Applications, Setubal,
Portugal, 26–28 July 2017; pp. 163–170.
Pires, I.M.S. Multi-Sensor Data Fusion in Mobile Devices for the Identification of Activities of Daily Living.
Ph.D. Thesis, Universidade da Beira Interior, Covilha, Portugal, 2018.
Pires, I.M.; Marques, G.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F.; Spinsante, S.; Teixeira, M.C.;
Zdravevski, E. Recognition of Activities of Daily Living and Environments Using Acoustic Sensors
Embedded on Mobile Devices. Electronics 2019, 8, 1499. [CrossRef]
Zdravevski, E.; Lameski, P.; Mingov, R.; Kulakov, A.; Gjorgjevikj, D. Robust histogram-based feature
engineering of time series data. In Proceedings of the 2015 Federated Conference on Computer Science and
Information Systems (FedCSIS), Lodz, Poland, 13–16 September 2015; pp. 381–388.
Zdravevski, E.; Risteska-Stojkoska, B.; Standl, M.; Schulz, H. Automatic machine-learning based
identification of jogging periods from accelerometer measurements of adolescents under field conditions.
PLoS ONE 2017, 12, e0184216. [CrossRef]
Lameski, P.; Zdravevski, E.; Mingov, R.; Kulakov, A. SVM Parameter Tuning with Grid Search and Its Impact
on Reduction of Model Over-fitting. In Rough Sets, Fuzzy Sets, Data Mining, and Granular Computing; Springer:
Cham, Switzerland, 2015; pp. 464–474.
Kelarev, A.V.; Stranieri, A.; Yearwood, J.L.; Jelinek, H.F. Empirical study of decision trees and ensemble
classifiers for monitoring of diabetes patients in pervasive healthcare. In Proceedings of the 2012 15th
International Conference on Network-Based Information Systems, Melbourne, Australia, 26–28 September
2012; pp. 441–446.
Electronics 2020, 9, 192
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
16 of 16
Xu, J.Y.; Chang, H.I.; Chien, C.; Kaiser, W.J.; Pottie, G.J. Context-driven, prescription-based personal activity
classification: Methodology, architecture, and end-to-end implementation. IEEE J. Biomed. Health Inform.
2013, 18, 1015–1025. [CrossRef]
Wiśniewski, M.; Zieliński, T.P. Joint application of audio spectral envelope and tonality index in an e-asthma
monitoring system. IEEE J. Biomed. Health Inform. 2014, 19, 1009–1018. [CrossRef]
Zhou, L.; Kang, Y.; Zhang, D.; Lai, J. Harmonized authentication based on ThumbStroke dynamics on touch
screen mobile phones. Decis. Support Syst. 2016, 92, 14–24. [CrossRef]
El Masri, A.; Wechsler, H.; Likarish, P.; Grayson, C.; Pu, C.; Al-Arayed, D.; Kang, B.B. Active authentication
using scrolling behaviors. In Proceedings of the 2015 6th International Conference on Information and
Communication Systems (ICICS), Amman, Jordan, 7–9 April 2015; pp. 257–262.
Xu, J.; Song, L.; Xu, J.Y.; Pottie, G.J.; Van Der Schaar, M. Personalized active learning for activity classification
using wireless wearable sensors. IEEE J. Sel. Top. Signal Process. 2016, 10, 865–876. [CrossRef]
Tang, C.; Phoha, V.V. An empirical evaluation of activities and classifiers for user identification on
smartphones. In Proceedings of the 2016 IEEE 8th International Conference on Biometrics Theory,
Applications and Systems (BTAS), Niagara Falls, NY, USA, 6–9 September 2016.
Yanyun, G.; Fang, Z.; Shaomeng, C.; Haiyong, L. A convolutional neural networks based transportation
mode identification algorithm. In Proceedings of the 2017 International Conference on Indoor Positioning
and Indoor Navigation (IPIN), Sapporo, Japan, 18–21 September 2017.
Li, S.; Qin, Z.; Song, H.; Si, C.; Sun, B.; Yang, X.; Zhang, R. A lightweight and aggregated system for
indoor/outdoor detection using smart devices. Future Gener. Comput. Syst. 2017. [CrossRef]
Qin, Y.; Jiang, M.; Yuan, W.; Chen, S.; Luo, H. Transportation mode recognition algorithm based on Bayesian
voting. In Proceedings of the 2017 5th International Conference on Enterprise Systems (ES), Beijing, China,
22–24 September 2017; pp. 260–269.
Vafeiadis, T.; Zikos, S.; Stavropoulos, G.; Ioannidis, D.; Krinidis, S.; Tzovaras, D.; Moustakas, K. Machine
learning based occupancy detection via the use of smart meters. In Proceedings of the 2017 International
Symposium on Computer Science and Intelligent Controls (ISCSIC), Budapest, Hungary, 20–22 October
2017; pp. 6–12.
Subasi, A.; Dammas, D.H.; Alghamdi, R.D.; Makawi, R.A.; Albiety, E.A.; Brahimi, T.; Sarirete, A. Sensor Based
Human Activity Recognition Using AdaBoost Ensemble Classifier. Procedia Comput. Sci. 2018, 140, 104–111.
[CrossRef]
Yuan, Y.; Melching, C.; Yuan, Y.; Hogrefe, D. Multi-device fusion for enhanced contextual awareness of
localization in indoor environments. IEEE Access 2018, 6, 7422–7431. [CrossRef]
Pires, I.M.; Garcia, N.M.; Pombo, N.; Flórez-Revuelta, F. Limitations of the Use of Mobile Devices and Smart
Environments for the Monitoring of Ageing People. In Proceedings of the ICT4AWE, Funchal, Portugal,
22–23 March 2018; pp. 269–275.
Pires, I.; Felizardo, V.; Pombo, N.; Garcia, N.M. Limitations of energy expenditure calculation based on a
mobile phone accelerometer. In Proceedings of the 2017 International Conference on High Performance
Computing & Simulation (HPCS), Genoa, Italy, 17–21 July 2017; pp. 124–127.
c 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).