Evaluation of Smartphone Inertial Sensor

Performance For Cross-Platform Mobile Appliances

Abstract:
Smartphone sensors are massively used in mobile applications. The working of the
mobile sensor changes from device to device as per the model and development of the
device. The cross-platform mobile application may be a complex and demanding task for
developers to help the publicly approachable mediums containing real-life situations
smartphone sensor boundary. To deal with this issue, we designed the participatory pilot
application and implemented this for measuring, collecting, analyzing the smartphone sensor
variables. They start up with the smartphone gyroscope and accelerometer noise and bias
parameter. It is an advanced but significant start, providing information on the measured
smartphonesensor’s statistical variables and insights into their performance. The second
step of this is the cloud base variable version as well the larger scale of this application is
already planned.
Introduction:
In the software market, the mobile application is an essential part. Also, the competition in
the mobile market is going to increase every year. The requirement for speedy application
development and distribution has never been greater. Due to the availability of many different
platforms and their dedication to the single platform software classification kit, the
competition is becoming more and more challenging in developing mobile applications. The
development concept in the single platform is an application id first classified for the specific
platform and then further developed for other platforms. His concept development provides
the path to the single platform application development, where an application id continuously
thrives for several platforms. To reduce the cost of cross-platform development and access
as many users as possible, the developer transfers to cross-platform development tools. As
per the discussion, these tools are more and more significant for upcoming years.
In recent years, the largest part of the mobile application that has been developed is primarily
for the smartphone. Today, smartphones include different software. The sensors in the
smartphones are based on the type or the model of the smartphone, proximity sensors,
camera as well the light sensor, temperature, and pressure sensors, etc. the plenty of the
different sensor types as well the sensor count heterogeneity of smartphone models are
posing the challenge for the process of development of cross-platform applications. The
second challenge is the sensor type varies from mobile to mobile models. The targeted
sensor must be workable at any targeted platform of any mobile device. The developers are
fully aware of the condition, or the parameter of the mobile device or the sensor parameter is
whether it is workable on any platform. This can access the mobile applications directly or a
built-in physical sensor. The raw sensor signals from physical sensors are proceeds and
made available to the application systematically as a smartphone sensor through the mobile
operating system. The main focus of this topic is the gyroscope and accelerometer sensors.
Both sensors are MEMS micro-electro-mechanical-system sensors. The MEMS is the
smartphone's physically embedded sensor. The application only accesses the mobile data
through the MEMS chip, not the sensor data, and the developer must consider this term.
Therefore the sensor parameters are not from the MEMS spad sheet.
Smartphone sensors are being used massively in growing mobile phone applications. The
mobile phone sensor is an application playing a significant role in our lives, such as fitness,
health, sports, and many other factors.
The smartphone sensing applications are used at many scales and prepared for multiple
purposes. Personal sensing applications are mainly designed for an individual user, with
recreation and well-being applications. Personal sensors applications doest share data with
others. The second one is the group sensor applications, and this is for the limited number of
audiences with the same aims, e.g., monitoring the garbage in the neighbourhood. In a
community sense, the application involves a large number of audiences and also collections
or sharing the data in bulk. The main problem with the sensing application was to what extent
these applications show the participant's activeness and resolve this issue. We discuss two
factors, (1) in participatory sensing, where the person is actively engaged in data collection,
(2) Opportunistic sensing involves the automatic system of data collection done by the
application automatically.
Mobile sensor devices provide many degrees of sensing quality. Some applications are not
able to provide high-quality data. The mobile sensing application fits perfectly in the IoE
Internet of Paradigm and the IoT Internet of things. But the IoT has limited working sensing
ability to a few quantities; regarding this, smartphones are less limited because cell phones
involve multiple sensors. Mobile phone applications are used in our daily life. Shortly, there
are many chances to grow in eHealth and eCARE. The good part of the research is that
logarithms tell about a patient's condition under observation. Sometimes, life depends on the
right and clear detection of their state. So the developers must create advanced algorithms to
avoid adverse events.
We agree that the mobile phone sensors greatly helped the research and application
development community. The researchers and the developers must know about the
accuracy of their data, especially when dealing with the eHealth application.
Motivation, Contribution, and Related Work:
Our research community is developing and studying the biochemical biofeedback system
and applications development based on initial sensors. Different feedback applications have
different demands regarding the accuracy, measurement, and precision. The boundary
values of parameters show the demands of this parameter and the errors induced by the
imprecision and inaccuracy. Many factors cause the error, but bias and noise are two. The
error source prompts the parameter value error that depends on the analysis time. e.gThe
gyroscope prompt angular error has the linear dependency on time and position, while the
accelerometer has the quadratic dependency on time.
The primary motivation is based on experience. The experience from the biofeedback
application and implementing the mobile sensing applications enable the mobile sensing
analysis measurement and storage sensors. Participatory sensing allows many
measurements in less time, which is the method of choice.
There are many results and benefits of mobile sensing applications. They provide accurate
and well-researched academic questions and solve real-world problems and issues
connected to the usage and development of cross-platform applications. Cross-platform
developers can take advantage of the statistical approach of the measured sensor on which
the person can draw the availability and the use of the application sensor for multiple
purposes, conditions, and demands.
● Compare the smartphones of the same platform.
● Compare the smartphone of the same manufacturer.
● Using different application comparisons of the model of the smartphone
● Comparison of the smartphone with the same physical sensor but a different platform.
● Compare the results having the same physical sensing in the same duration of time.
A possible difficulty for the successful implementation of the mobile sensing application is to
need a large audience or participants. Mainly the mobile applications start the small project
that further expands to the large scale. The small group of participants involves tens or
hundreds of people, most commonly the students in academic applications. Fo the large
scale needed for our mobile application, we focus on the incentive that will help in the
mechanism of source work, (1) participation in research programs, (2) quick sensor quality
information that sent to the user or participant, absolute value and the co person of the value
with other participants, (3) gamification.
The other hurdle also deals with the measurements that deviate from the protocol.
● The usefulness of the participatory data and its confirmation accession concept for
measurement and assembly of smartphone sensor parameters This is gained
through the implementation, development, and employment.
● Anthology of the database contains the measured sensor parameters of almost 60
smartphones. This is gained through the recruitment of almost one hundred
participants, and the participants collectively performed more than 500
measurements of their smartphone sensors by using the application pilot
participatory.
● Useful findings and the results about the smartphone's properties and their statistical
approach can be used as the base for the developers of the smartphone cross-
platform application.
Related Work:
Participatory sensing applications are designed to sense the physical quantities of interest,
including air pollution, temperature, body activity, etc. In smartphones, these application
sensors measure the quantity of interest. The main focus is gathering the data, and the least
focus is on sensing data quality. This cannot use the raw smartphone in participatory sensing
applications in high demand. This may have unbearable errors that must be reduced before
using the sensor data in sensing applications. The smartphone sensory application for air
quality measurement needs high data quality. It measures the smartphone sensing readings
compared to the government data for the proper results. The sensor data quality developers
cross platforms must coincide with the high variability of the data sensing parameters. The
issues regarding the data variability of the sensing smartphone sensors are based on the
measurement data from the nine different smartphone models. Several researchers take part
in the evaluation and calibration of smartphone data sensing. The author performed the
laboratory measurements of the gyroscope and accelerometer performance for three high-
end mobile devices, the autonomous calibrations based on the sensor fusion. The two
smartphones considered the computation and time scale of the gyroscope and
accelerometer measured.
Pilot Participatory Sensing System:
The executed pilot system employs the participatory sensing idea, where the participants
actively participate in the data collection. The model of this pilot participatory sensing system
is described through the diagram. There is a connection between the participant's
smartphone and one available wireless and sends data to the network internet server. The
servers extract the data information, analyze it, and then write it on the database; this results
from the information. The participants allow using the database to check the results
individually. They mainly developed the pilot system for the group of modern smartphone
participants.
Figure 1. Pilot system architecture. Smartphones send sensor data over one of the available
wireless interfaces to the processing computer (server). Processing results are stored in the
database. They can be analyzed and retrieved through a database access application.
In the pilot implementation, a member first needs to install the application setup, one of the
reinforced shelf applications, to stream sensor data to the distant location. For this purpose,
we use three different applications
● Sensor Node Version 1.53 for the Android platform
● Sensor stream version 1.1 for the iOS
● sensor monitor pro version 1.0.9
Each set of the sensors measuring episodes necessarily be actively started by the
participant. By starting the measuring episode, sensor data is sent to the server and
analyzed. Results are written on a database—the working of a server in the public IP network
and running the customer-designed view LaB VIEW applications. The database is running on
Windows 2008 R2 like XAMPP and MySQL.
The pilot implementation was massively used by the students, their colleagues, and the
engineering students.
Measurement Protocol:
The main focus is to measure a limited number of smartphones with gyroscope or
accelerometer parameters. The measurements are held and performed at different subjects
in different locations through the participatory sensing parameters. The protocol for the pilot
participatory sensing application must be straightforward. e.gThere is no need for hard-to-get
tools, which could perform in a lenient environment; each measurement should not take too
long. The limitation proposes comparing the complexity of the measurement protocols and
the quality of measurement. Therefore the looser measurements gave loose results that are
low quality. Therefore only choose the set of the limited smartphone that gave enough
reliability under the given protocol limitations. We measure the limited noise measurements
for smartphone gyroscope or accelerator for smartphone gyroscope or accelerator.
The smartphones and n still-one positions perform the noise and sensor bias measurements.
The accelerometer measurement needs a leveled flat area. All the user's advised to use all
the spirits to measure the correct surface. The accuracy of the surface measures values lies
between the 0.5mm/m and the 1mm/m. Due to the gravity projection, the measured value
parallel to the surface is 1mg. This is necessary. The smartphone must have good
connections. In every measurement, the inaugurated sensor data streaming, in addition, the
smartphone put on the flat surface face down. Since many smartphones have a flat surface,
the position determines the smartphone's orientation. The MEMS sensor is aligned with the
screen. The z-axis lies perpendicular to the screen. This means that gravity is not affected by
the X-axis and Y-axis of the accelerometer. This ensures that the smartphone does not have
high vibrations; every measurement takes 100s.
Measurement Methodology:
The smartphone sensor imprecision and inaccuracy limit the value of the smartphone sensor.
Sensor bias and the noise cause values that produce the linear, angular value of quadratic
and gyroscope position of error accelerometer.
Definition of the sensor bias as an average sensor output at zero sensor input. The bias
value equation is (1) is approximate by averaging N samples of sensor signals and times
depending on the averaging frequency fs and signal sample block length denoted by N.
N−1
1
x bias=
N ∑ ❑ x[n] (1)
n =0
By measuring the Allan Variance sensor, noise value can be determined. Equation (3) is
defined as Allan variance. This is defined as the average variance bias value estimated bias
value y[m] is calculated from a block N signal sample.
N−1
1
y[m]=
N ∑ ❑ x[n+m.N] (2)
n =0

2 1
σ A [N]= (y[m]- y[m-1])❑2 (3)
2

Allen variance is defined from a fine data stream in value M successive bias determined y[m]
in equation 3,
m −1
1
σ 2A ≈ ∑ ❑(y[m]-y[m-1])❑2
2( M −1) m =1

In equation 4, an approximate error occurs,

1
δ σ=
√2( M −1)
Allen variance is the function of block length N; the second method of expressing this is
averaging time T ave =N /f s. The noise in the sensor is generated from different random
processes. Many noise terms with different spectral line densities occur simultaneously. If the
average time is around 1s, the white noise is the common error source for MEMS
accelerometer and gyroscope. In the condition the Allen variance measures in average time
t ave =1 s t h at represent t h e sensor w h ite noise density . Velocity random walk (VRW) for the
accelerometer and the gyroscope(GRW) parameter nσA at Tag = 1 s Allen deviations.
Allan variance functions are commonly shown in the graphs. Its calculations need long
periods, long measurement time, and the number of sample signals. For accurate results, the
measurement time must be greater than ten times compared to the average time to present
the function graph.
Results:
In implementing a sensory pilot application, we collected the data almost from 116
smartphone devices, and the measurement recorded was 500 in just 44 days. The
smartphone collection involves the 61 models from which the 31 devices measured once and
the remaining devices measured twice. The mobile devices that measure more than once
include GalaxyS3, Galaxy S4, iPhone 4, iPhone 5, iPhonr5s, and Xperia Z1 compact. The
measurement smartphone was from 13 different manufacturers and used two different
platforms. The measurement results are,
● Complete dataset results
● Results by smartphone models
● Result of single smartphone device.
Complete Dataset Results:
The measuring devices mention the complete dataset in the section we deal with the
measured data. Plot the average value parameter from the smartphone. The plots show the
smartphone gyroscope bias and accelerometer, their average accuracy.

Figure.2 (a)

Figure2(b)
Figure 2. Accelerometer and gyroscope bias measurements of X, Y, and Z axes.
● (a) Average accelerometer biases of 116 different smartphones are plotted;
 ● (b) Average gyroscope biases of smartphones with gyroscopes are plotted. The
horizontal axis represents the device identification number (ID).

In the figure, we measure the accelerometer and gyroscope measurements of the X, Y, and
Z, axis. The horizontal surface shows the identification of the device for the database.
● The complete accelerometer measurement dataset is plotted, including an example
of a faulty device with ID = 50 that deviates from other devices. The vertical axis
shows VRW in [g0/ Hz ]
● The complete gyroscope measurement dataset is plotted, including the examples of
the three faulty devices with ID = {25, 50, 73}. The vertical axis shows ARE in [deg/s/
Hz ].
Accelerometer
(a) and gyroscope
(b) the noise of all properly functioning devices (Ka = 108, Kg = 82). The horizontal axis
represents the device identification number from the measurement database. Considerable
deviations result from different models of MEMS sensor chips embedded in different
smartphone models.
The vertical axis shows
(a) VRW in [g0/ Hz ]
(b) ARW in [deg/s/ Hz ].
Table1: statistical values of the measured parameters, the gyroscope, and accelerometer
bias. Standard deviation accuracy and a few percentiles are mentioned in the table.

Accelerometer [mg0] Gyroscope [mrad/s]

Parameter X Y Z X Y Z
Average 14.3 14.6 25.3 9.4 8.7 6.1
StDev 14.2 15.2 25.1 13.6 12.1 8.7
50th percentile 10.0 9.9 18.5 3.1 4.3 2.8
90th percentile 30.1 31.5 60.3 30.8 22.9 17.1
95th percentile 43.6 45.9 71.1 40.5 35.4 28.2
100th percentile 90.9 82.7 161.0 142.7 81.7 158.2
Figure 2(A) averaged accelerometer bias from all the devices. Also, figure (2)B represents
the gyroscope bias averaged value. One device has a larger bias, but this is not the cause of
the fault. e.g., The excessive value of gyroscope bias in figure 2 B shows the evident results
from the ID= 73 in figure 2b. The gyroscope and accelerometer should measure and
compensate in many applications. The low competition application can work without bias.
The statistical data is collected in the table1. The measured bias value of 50% of the device
average below the 95% percentile might be helpful for the developers in cross-platform
applications. For the more accurate results require the cloud-based application full-scale.
The VRW & ARW are measured under the fully defined single-point Allen variation
measurements. A larger deviation could be helpful to measure the faulty diagnoses in the
devices.
The smartphone having ID=50 from the VRW causes the nose parameter and the ARW the
device ID=25,50,70 massive gyroscopic density nose parameter as the results of these
devices move from the further analysis. Figure four shows the results for larger deviation
results. The sensor embeds in the device in large deviations and counts every sight moment
or noise.
Results by Smartphone model:
While the statistical analysis gives us the complete picture of measurement of dataset
smartphone statice, this is very important to find out the statistical data of the particular
smartphone model. In these statistics, the model of different smartphones is compared. The
selection of smartphone devices includes Galaxy S3, Galaxy S4, iPhone4, iPhone5,
iPhone5s, iPhone6, Xperia Z1, Nexus 5. In the figure, 5’6 used the same set of mobiles
models.
VRV and ARW are the noise parameter values that stick with the embedded MEMS chip. We
set the devices separately for the dataset in figure 4, the smartphone model for the
discussion. In this dataset measurement, use six different devices with the same models.
Figure 5 represents the comparison of the average accelerometer noise density and VRV
combined with its standard deviation. Furthermore, figure 6 represents the gyroscope ARV
noise density compared to the standard deviation.
The results of 5’6 follow the random order. A similar difference in noise parameters between
various phone devices is evident from the principal axis. The mobile devices 4’8 show the
best results sensors. The smartphones 1’2’3 show below average for gyroscope and average
for the accelerometer. While device 7 has a low average for the accelerometer and an
average for the gyroscope.

Figure 5. Comparison of average accelerometer noise parameter VRW and its standard
deviation for eight different smartphone models. Smartphone models presented are Galaxy
S3, Galaxy S4, iPhone 4, iPhone 5, iPhone 5S, iPhone 6, Nexus 5, Xperia Z1 Compact. The
listed smartphone models are presented in a random order (the same for all graphs). The
vertical axis shows VRW in [g0/ Hz ].
Figure 6. Compare the average gyroscope noise parameter ARW and its standard deviation
for eight smartphone models. The smartphone models presented are the same as in Figure
5. The vertical axis shows ARE in [deg/s/ Hz ].

Result by Individual Device:

The results and the complete dataset module are in 4.1 and 4.2, which would be helpful for
the cross-platform developers. Every participant may use this helpful when comparing wh the
smartphone devices with all measuring devices. But the users who measure their devices
after some time can benefit from the statistical data of their own mobile devices. Figure 7
represents that when the user checks his phone 44 times, the table 2’3 indicates the mobile
devices viewed in the 44 days measured more than ten times.
Many measurements can give information about the important sensor parameter for the
same sensor device, including bias variation.
The gyroscope bias and accelerometer vary with time. Bias variations result from the random
low-frequency noise and depend on the temperature variations; without measuring the
variation in temperature, we cannot set the deterministic bias drift.
Many smartphones are repeatedly measured in 44 days. Picture 7a,d shows the result of
N=44 accelerometer and gyroscope bias measurements involving the iPhone 4s with ID=3.
For 44 days, this device was under observation every evening. There are significant

variations in the bias measurements in scale every day.

Figure 7. Repetitive accelerometer and gyroscope bias measurements (N = 44) of the
smartphone with ID = 3 showing bias variation. Measurement numbers are taken from the
database and are not successive as other measurements taken between two measurements
of the presented device. Accelerometer bias is in [g0], gyroscope bias is in [rad/s].
For bias variation from the different mobile devices, the different smartphones were selected,
Xperia Z1 Compact ID=2, iPhone 4 ID=3, iPhone 5s ID=7, iPhone 6s ID=8, and the LG
Nexus ID=14.
Table 2 represents the measurements for the accelerometer bias, and in table 3, the bias
variation is not critical for all devices under observation.

Table 2. Statistical parameters of smartphone accelerometer bias. Selected devices with

more than ten measurements are presented. Average and standard deviation [g0] for the
three axes are listed.

Average S.D Min-Max

ID N X Y Z X Y Z X Y Z

2 41 0.0063 0.0279 0.0231 0.0057 0.0019 0.0018 0.0236 0.0065 0.0075

3 44 -0.0121 -0.0110 -0.0264 0.0005 0.0010 0.0038 0.0030 0.0040 0.0171
7 42 0.0051 0.0060 0.0034 0.0009 0.0012 0.0011 0.0045 0.0046 0.0056
8 14 0.0029 0.0032 0.0004 0.0025 0.0009 0.0008 0.0097 0.0025 0.0030
14 11 -0.0241 -0.0259 -0.0994 0.0006 0.0017 0.0013 0.0058 0.0058 0.0046

Table 3. Statistical parameters of smartphone gyroscopes bias. Selected devices with more
than ten measurements are presented. Average and standard deviation [rad/s] for the three
axes are listed.
 Average S.D Max-Min

ID N X Y Z X Y Z X Y Z

2 41 0.0000 0.00002 0.00001 0.00013 0.00009 0.00010 0.00051 0.00042 0.0045

3 44 -0.01495 0.00833 0.00423 0.00158 0.00091 0.00116 0.00591 0.00268
0.00380
7 42 0.05062 0.05244 0.01190 0.00396 0.00054 0.00133 0.01659 0.00221 0.00812
8 14 0.002728 0.03175 -0.00258 0.00054 0.00134 0.00037 0.00161 0.00410 0.0013
14 11 -0.00004 0.00002 0.00001 0.00016 0.00036 0.00013 0.00052 0.00142
0.00039

Discussion and Future Work:

The results of the process are interesting and encouraging. Even in the limited criteria, these
are the uses for the help of developers in cross-platform applications. A larger-scale provides
better information about the smartphone's sensor applications. The results of this would be
helpful in further research, like the maintenance of the data and the development of the cross
platforms sensing applications.
The next development step is to start up cloud bias sensing cross-platform applications. The
application would operate with the clouds to analyze the person's data. The development of
cloud applications is a complex task. The concept Sasa means Such as Sensing services
has been a purpose to making the task easy to design mobile sensing applications. The next
plan up starting principle is described in figure 8. This system is the same as many other
system applications. The system uses the employee's paradigm, where the members
actively work to collect the data. The first step is to install the application to develop a
sensing client onto the mobile phone. Each measuring startup actively participated. After
setting the measuring episode, the data is sent to the processing instance cloud. Then the
processing data is written on the cloud. For the data retrieval, a web application is used from
the database and its statistics. Participants can find their results from the processing unit
immediately after measuring the episode.
Figure 8. Cloud-based participatory sensing system architecture. Smartphones send sensor
data to the cloud for processing. Processed sensor data results are stored in the cloud
database. This can retrieve the results directly from the processing unit or a cloud web
application.
The system is upgraded for other functionalities except for sensor performance
measurements and data analysis in the long run. e.gThe less demanding feedback can use
the existing module or infrastructure to send the data processing unit; the processing unit
processes the data and then sends it back to the participant. Furthermore, the system
functionalized the opportunity functionality by taking measurements regularly with the
previous consent of the participant at the favorable time. Not all the sensors are measurable
through the opportunistic paradigm.
Privacy is the main concern after collecting the data from the participants. Also, the sensing
participatory application provides enough measures to avoid any issue. The purpose of
presenting an anonymization system works inside the installed phone application. The
smartphone ID hash will only be known as the participating initiating measuring episode.
The same hash is again used when the participant knows the results from the application
cloud.
Conclusion:
Do you know how good your smartphone sensor is? Remember that smartphones do not
show directly reflected embedded parameters; measuring the main or former is significant.
The collected participatory sensing applications show no direct answer to the above
question, and the parameter measured varies from the smartphone model and sometimes
varies in the same model of smartphone. By analyzing the measurement data, we learned
that results vary in the same mobile phone models, nose varies in the mobile devices, and
same for the same model.
How can the developers of the cross-platform benefit from the results? Knowing that the
different applications demand different sensor qualities could reduce the noise, and the bias
can be compensated but not eliminated, but it all depends on the developer's approach. For
example, if the demand for the application is low, there is no need to compensate the basis
at all. It is providing that the targeted amount of research and group of smartphones exhibits
the bias that fulfills the application requirement.
The data of the pilot application give an overview of the statistics of the smartphone
accelerometer and gyroscope data. To improve the relevance of the resulting increase in the
number of participants, platforms, and sensor parameters would improve the statistical
relevance of the result. There are many ideas under consideration to improve the
participatory sensing applications. The top priority of the list is to upgrade the code bise
applications.
Acknowledgments:
The Slovenian agency supported this research work within the algorithms of research
programs, the optimization methods in telecommunications, the use of projected wireless
devices for Biofeedback, and motion analysis. Publish costs are covered in the research
programs' algorithms and optimization methods in telecommunications.
Author Contribution:
All authors took part in this article significantly. The article is structured and convinced by all
the authors. Anton kos and Anton Umek performed the expertise data analysis and wrote the
paper.
Conflicts of interests:
There are no conflicts of authors declared by the author. The founding sponsor has no role in
the study, interpretation of data, or analysis.
References:
● Tilak, S. Real-world deployments of participatory sensing applications: Current trends
and future directions. ISRN Sens. Netw. 2013, doi:10.1155/2013/583165.
● Saeedi, S.; El-Sheimy, N. Activity Recognition Using Fusion of Low-Cost Sensors on
a Smartphone for Mobile Navigation Application. Micromachines 2015, 6, 1100–1134.
● Edmond, M.; Monaghan, D.; O’Connor, N.E. Classification of sporting activities using
smartphone accelerometers. Sensors 2013, 13, 5317–5337
● McGrath, M.J.; Scanaill, C.N. Wellness, fitness, and lifestyle sensing applications. In
Sensor Technologies; Apress: New York, NY, USA, 2013; pp. 217–248. 7.
● Lane, N.D.; Milazzo, E.; Lu, H.; Peebles, D.; Choudhury, T.; Campbell, A.T. A survey
of mobile phone sensing. IEEE Commun. Mag. 2010, 48, 140–150.
● Moschino-Herranz, I.; Gil-Pita, R.; Ferreira, J.; Rosa-Zurera, M.; Seoane, F.
Assessment of Mental, Emotional and Physical Stress through Analysis of
Physiological Signals Using Smartphones. Sensors 2015, 15, 25607–25627.
● Kos, A.; Tomažič, S.; Umek, A. Suitability of smartphone inertial sensors for real-time
biofeedback applications. Sensors 2016, 16, doi:10.3390/s16030301
● Gupta, M.; Holloway, C.; Heravi, B.M.; Hailes, S. A comparison between smartphone
sensors and bespoke sensor devices for wheelchair accessibility studies. In
Proceedings of the 2015 IEEE Tenth International Conference on Intelligent Sensors,
Sensor Networks and Information Processing (ISSNIP), Singapore, 7–9 April 2015;
pp. 1–6.
● Feng, M.; Fukuda, Y.; Mizuta, M.; Ozer, E. Citizen Sensors for SHM: Use of
Accelerometer Data from Smartphones. Sensors 2015, 15, 2980–2998.
● Ganti, R.K.; Ye, F.; Lei, H. Mobile crowdsensing: Current state and future challenges.
IEEE Commun. Mag. 2011, 49, 32–39.
● Casilari, E.; Luque, R.; Morón, M-J. Analysis of android device-based solutions for
fall detection. Sensors 2015, 15, 17827–17894.
● Sheng, X.; Xiao, X.; Tang, J.; Xue, G. Sensing as a service: A cloud computing
system for mobile phone sensing. In Proceedings of the 2012 IEEE Sensors, Taipei,
Taiwan, 28–31 October 2012; pp. 1–4.
● Hasenfratz, D.; Saukh, O.; Sturzenegger, S.; Thiele, L. Participatory air pollution
monitoring using smartphones. In Proceedings of the 2nd International Workshop on
Mobile Sensing, Beijing, China, 16–20 April 2012.
● Stančin, S.; Tomažič, S. Time-and Computation-Efficient Calibration of MEMS 3D
Accelerometers and Gyroscopes. Sensors 2014, 14, 14885–14915.
● Grewal, M.; Andrews, A. How good is your Gyro. IEEE Control Syst. Mag. 2010, 30,
12–86.
● Weinberg, H. Gyro Mechanical Performance: The Most Important Parameter;
Technical Article MS-2158; Analog Devices: Norwood, MA, USA, 2011.
● El-Shamy, N.; Hou H.; Niu, X. Analysis and modeling of inertial sensors using Allan
variance. IEEE Trans. Instrum. Meas. 2008, 57, 140–149.
● Aggarwal, P.; Syed, Z.; Niu, X.; El-Sheimy, N. A standard testing and calibration
procedure for low-cost MEMS inertial sensors and units. J. Navig. 2008, 61, 323–336.
● El-Dusty, M.; Pagiatakis, S. Calibration and stochastic modeling of inertial navigation
sensor errors. J. GPS 2008, 7, 170–182.
● Das, T.; Mohan, P.; Padmanabhan, V.N.; Ramjee, R.; Sharma, A. PRISM: Platform
for remote sensing using smartphones. In Proceedings of the 8th International
Conference on Mobile Systems, Applications, and Services, ACM, San Francisco,
CA, USA, 15–18 June 2010; pp. 63–76.
● Ozer, E.; Feng, M.Q.; Feng, D. Citizen Sensors for SHM: Towards a Crowdsourcing
Platform. Sensors 2015, 15, 14591–14614.
● Bouts, I.; Kalogeraki, V. Privacy preservation for participatory sensing data. In
Proceedings of the 2013 IEEE International Conference on Pervasive Computing and
Communications (PerCom), San Diego, CA, USA, 18–22 March 2013; pp. 103–113.