0% found this document useful (0 votes)
72 views5 pages

Sensors in Mobile Devices Knowledge Base

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
72 views5 pages

Sensors in Mobile Devices Knowledge Base

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/339396270

Sensors in Mobile Devices Knowledge Base

Article  in  IEEE Sensors Letters · February 2020


DOI: 10.1109/LSENS.2020.2975161

CITATIONS READS
3 436

3 authors:

Igor Khokhlov L. Reznik


Rochester Institute of Technology Rochester Institute of Technology
23 PUBLICATIONS   83 CITATIONS    133 PUBLICATIONS   994 CITATIONS   

SEE PROFILE SEE PROFILE

Sahil Ajmera
Rochester Institute of Technology
1 PUBLICATION   3 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Data quality and security evaluation for mobile devices View project

All content following this page was uploaded by Igor Khokhlov on 23 September 2020.

The user has requested enhancement of the downloaded file.


VOL. 4, NO. 3, MARCH 2020 5500404

Sensor systems

Sensors in Mobile Devices Knowledge Base


Igor Khokhlov∗ , Leon Reznik∗∗ , and Sahil Ajmera
Golisano College of Computing and Information Science, Rochester Institute of Technology, Rochester, NY 14623 USA
∗ Student Member, IEEE
∗∗ Senior Member, IEEE

Manuscript received January 22, 2020; revised February 11, 2020; accepted February 17, 2020. Date of publication February 20, 2020; date of current
version March 10, 2020.

Abstract—Multiple sensors are incorporated in modern mobile devices. The ubiquity of these mobile devices with a wide
range of embedded sensors enables developing new types of sensor networks and novel applications. However, more
often than not, users are unaware of the sensor quality in their devices. With the goal of facilitating closer collaboration
between sensor developers and users’ communities, we developed a knowledge base that contains various characteristics
of sensors embedded into mobile devices and their quality analysis. In our novel analysis, we fused both sensor and
mobile device characteristics to produce integral sensor quality indicators. Machine learning and clustering techniques
were employed. To validate our approach, we collected comprehensive information about sensors currently embedded in
more than 9000 mobile devices and analyzed them. We made this knowledge base and evaluation results available to the
community. Also, an Android OS-based application that evaluates the quality of sensors in a particular device and provides
recommendations on their use was developed and presented for community use. Examples of possible knowledge base
applications are given.

Index Terms—Sensor systems, mobile devices, sensor quality.

I. INTRODUCTION This research aims at building up the knowledge base that includes
major characteristics of sensors currently incorporated into smart-
With multiple sensors incorporated into popular mobile commu- phones and analyzes them with the goal of evaluating the overall
nication devices, such as smartphones, their owners have the ability quality of data coming out. This knowledge base will become a valu-
to employ them in a growing number of crowdsourcing [1], [2] and able asset for both groups of sensor developers and users. Also, in this
crowdsensing applications [3], [4]. This sensor technology progress article, we present tools we have developed to facilitate collaboration
has turned smartphone owners into sensor users, with many of them between sensor developers and sensor users to make them aware of the
not even realizing it. The sensor community should embrace closer col- sensors, which the smartphone holders have, their characteristics, and
laboration between sensor developers and sensor users. Users should possible use. Sensor properties such as sensitivity, noise, resolution,
learn how and where to use their sensor instruments and what new nonlinearity, and cross-axis sensitivity determine the precision, by
sensor-based applications they can develop. This collaboration will which sensors sense the real world. With the help of crowdsourced data
result in expanding the sensor market and stimulating further sensor and information about the properties of sensors from various sensor
research and technology advancement. Extending and improving the manufacturers, we can provide valuable knowledge about the quality
sensor user’s pool may open up amazing opportunities in creating a and usage of sensors in various kinds of applications. This article
new sensor-based infrastructure that would allow employing a wider reports the results of the project, in which we studied characteristics
range of sensors, building more diverse sensor networks in order to of sensors that are built into mobile devices, analyzed and compared
improve the coverage, functionality, energy consumption, and overall them in order to estimate the quality level of these sensors, and
quality of data collection services given to the community. developed awareness and educational tools that included a smartphone
Unfortunately, many smartphone holders have very limited knowl- app, aimed at delivering this sensor knowledge to mobile device users
edge about embedded sensors and their characteristics, quality, and in order to improve their understanding about sensor possibilities and
constraints, in comparison to other instruments. The lack of expe- the limitations they have. As we estimate, our study covers over 60%
rience in dealing with sensors may confuse users in their choice of mobile device sensors available on the market.
of applications, where their contribution might be valuable. Even While we are developing a generic knowledge base approach, pre-
worse, not intelligent use of sensor-based applications available on vious studies and tools were highly specialized with a very limited
mobile devices may result in severe and dangerous violations of number of sensors and sensor types involved. Das [5], for example,
the user’s privacy, computer security, and the safety of the whole collected sensor data in real time under different conditions by creating
infrastructure. The sensor community should take the lead in advancing a web application and storing the data on a university server. Unfor-
sensor-related knowledge bases and developing tools to educate mobile tunately, since web applications can access only an accelerometer and
device owners about sensors and effective and secure ways to employ a gyroscope sensor of a mobile device, this application can be used
them. only with two mobile sensors. Hilgenberg [6] develops a configurable
and device-independent mobile sensor data collection framework that
provides sensor readings in real time. Ma et al. [7] compared ac-
Corresponding author: Igor Khokhlov (e-mail: [email protected]). celerometers, gyroscopes, and magnetometers based on properties of
Associate Editor: J. M. Corres.
Digital Object Identifier 10.1109/LSENS.2020.2975161
sensors, such as sensitivity, noise, frequency, and range. Mourcou

1949-307X © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 23,2020 at 21:18:27 UTC from IEEE Xplore. Restrictions apply.
5500404 VOL. 4, NO. 3, MARCH 2020

Table 1. Sensors and the Characteristics Upon Which They Were


Evaluated.

Fig. 1. Data collection and processing workflow.


Table 2. Statistical Analysis for the Accelerometer Sensor Type.
et al. [8] evaluated mobile device sensors for clinical motion research.
Shala and Rodriguez [9] compared the accuracy of mobile device
sensors for indoor positioning scenarios. Unlike the aforementioned
research, we develop a service that is applicable to a wide range of
applications and embedded sensors. In contrast to the research and
frameworks mentioned above, in our knowledge base, we consider
multiple sensor quality metrics, integrate them into the overall quality
Table 3. Statistical Analysis for the Gyroscope Sensor Type.
score, classify sensors into two quality groups using machine learning
techniques, and provide information about sensor quality to users
through the developed Android OS application.
The rest of this article is presented as follows. Section II presents
the sensor data collection and preprocessing procedures. In Section III,
we describe our sensor parameter analysis. The quality evaluation is
presented in Section IV. The collected sensor information is made
available as an open-source project and could be used by the com-
munity as we provide the sensor description and user guidance tools.
III. MOBILE SENSOR DATA COLLECTION
In order to facilitate a regular smartphone owner learning about the
sensors, to encourage their use, and to educate them, we produced Our data collection contains information about sensors embedded
an Android OS app that lists the sensors embedded in a particular in 9443 devices, including Android OS-based smartphones, wearable
device and evaluates their quality (see Section V for further details). devices, and tablets. This data collection includes other attributes re-
Section VI presents these project application examples. Finally, Sec- lated to mobile devices, such as type, size, resolution, chipset, camera,
tion VII concludes this letter. sensors, and performance. In total, we have about 58 attributes for
these devices produced by over 100 manufacturers.
The current collection contains data on 19 sensor types [17], in-
II. SENSOR DATA COLLECTION AND cluding accelerometer, barometer, gesture, humidity [18], face id and
PROCESSING METHODOLOGY fingerprint sensor (rear, side, under display, and in front), temperature,
ambient light, gyroscope, heart rate, compass, UV [19], iris scanner,
This research is focused on the most common sensors that are S po2 [20], proximity, and color spectrum sensors. In this research,
embedded in modern mobile devices. To obtain data about these we selected and further analyzed the five most popular sensor types:
sensors, we searched public repositories such as GSMArena [10] that 52 accelerometers brands, 14 gyroscopes brands, 20 proximity sensor
allowed us to acquire a comprehensive list of sensors and their types. brands, five barometers, and 16 compass sensor brands. The selected
Also, we investigated sensor characteristics that can be found in the sensors and their characteristics that are chosen for further analysis
manufacturers’ datasheets and online [11]–[15]. The data collection are given in Table 1. For each of the chosen sensor types, we selected
workflow is presented in Fig. 1. sensor parameters that can be accessed through the standard Android
We made the collected data publicly available, and we encourage the OS API. These parameter‘s values are used for the sensors’ quality
community to use it [16]. We further analyzed the original dataset and evaluation. The selected sensor parameters and the results of their
identified the attributes related to the selected sensors. We normalized statistical analysis are presented in Table 2 for accelerometer sensors,
sensor characteristics in the range between “−1” and “1.” In the case Table 3 for gyroscope sensors, Table 4 for proximity sensors, Table 5
of sensor quality metrics that have a positive influence on the overall for barometer sensors, and Table 6 for magnetometer sensors.
sensor quality score, “−1” represents the worst case, “0” is the average,
and “1” is the best case. For those sensor quality metrics that have
a negative influence on the overall sensor quality score, the scale is IV. MOBILE SENSOR QUALITY ANALYSIS
inverted, where “−1” represents the best case and “1” is the worst
case. This normalization facilitates further comparison of sensors and In our further analysis, we employed the K-means method to classify
their quality evaluation. all sensors of each type into groups based on their quality. The K-means

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 23,2020 at 21:18:27 UTC from IEEE Xplore. Restrictions apply.
VOL. 4, NO. 3, MARCH 2020 5500404

Table 4. Statistical Analysis for the Proximity Sensor Type. Table 7. Sensor Clustering Based on Sensor Quality Metrics.

Acc.: Accelerometer, Gyro.: Gyroscope, Prox.: Proximity, Bar.: Barom-


Table 5. Statistical Analysis for the Barometer Sensor Type. eter, Comp.: Compass.

Table 8. Device Classification Based on Sensors Quality.

Acc.: Accelerometer, Gyro.: Gyroscope, Prox.: Proximity,


Table 6. Statistical Analysis for the Compass Sensor Type. Bar.: Barometer, Comp.: Compass.

Fig. 3. Screenshot of the Android OS application for sensor quality


assessment.
Fig. 2. Quality score evaluation workflow.
on their sensitivity, nonlinearity, and noise values. Since, in the de-
veloped knowledge base, each sensor type has its own table, we do
clustering algorithm is chosen, as it is relatively computationally
not include sensor type in the table. Table 8 presents the results of
inexpensive and, considering small quality metrics dimensionality,
our device classification based on their embedded sensors’ quality. All
provides good cluster positioning [21]. This algorithm allows us to
devices were clustered based on one of the five described sensors; for
partition a collection of instances into k different subsets employing a
example, there are 119 devices with an accelerometer sensor of good
recursive technique. Since this algorithm requires choosing the initial
quality and 343 with a poor quality accelerometer.
(start) point of cluster centroids, there are several approaches to choose
them. In our research, we employed the K-means++ algorithm, which
uses a heuristic to find centroid seeds for k-means clustering, and V. SENSOR QUALITY EVALUATION TOOL
take advantage of our sensor quality metrics analysis and scaling.
According to Arthur and Vassilvitskii [22], K-means++ improves the To assist mobile device users in the quality analysis of sensors that
efficiency and effectiveness of Lloyd’s algorithm. are built into their devices and to provide further recommendations, we
The workflow for quality evaluation is presented in Fig. 2. Since not developed the specialized application [23]. It is based on the quality
all sensor parameters equally affect the overall sensor quality, in the calculation procedures described in the previous section and can be
quality evaluation calculus, weights to each of the sensor parameters executed on the Android OS devices. This application does not require
can be adjusted. In this particular implementation, we employed the any specific Android permissions and employs standard Android API
mean value of all metrics. One can see that metrics normalization in the to acquire and to present to the user sensor types that are built-in
range from “−1” to “1” facilitates the overall sensor quality evaluation, in a smartphone, as well as sensor characteristics. While basic sensor
since only weights have to be adjusted based on the metric importance. characteristics, such as the resolution or sensing range, may be directly
The overall quality indicators were calculated for each sensor type, and accessed through the Android API, other specific information such as
the results are made available online [16] as well. noise density or nonlinearity is acquired from the developed database.
Table 7 presents the results of the clustering of the accelerometers, The screenshot of the application’s user interface is presented in
gyroscopes, proximity sensors, magnetometers, and barometers based Fig. 3. The developed Android OS application reveals to the users those

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 23,2020 at 21:18:27 UTC from IEEE Xplore. Restrictions apply.
5500404 VOL. 4, NO. 3, MARCH 2020

sensors that are built-in in smartphones, their basic characteristics, and [2] A. Ghezzi, D. Gabelloni, A. Martini, and A. Natalicchio, “Crowdsourcing: A review
overall quality score. and suggestions for future research,” Int. J. Manage. Rev., vol. 20, no. 2, pp. 343–363,
2018.
[3] X. Zhang et al., “Incentives for mobile crowd sensing: A survey,” IEEE Commun.
Surv. Tut., vol. 18, no. 1, pp. 54–67, Jan.–Mar. 2016.
VI. KNOWLEDGE BASE APPLICATIONS [4] R. K. Ganti, F. Ye, and H. Lei, “Mobile crowdsensing: Current state and future
challenges,” IEEE Commun. Mag., vol. 49, no. 11, pp. 32–39, Nov. 2011.
The created knowledge base can be used in a variety of domains, [5] A. Das, Sensor Data Collection, 2019. [Online]. Available: https://www.cs.cmu.edu/
for example, in crowdsensing, in which it will allow choosing and anupamd/SensorDataCollection.html. Accessed: Nov. 9, 2019.
fusing the best sensor sources for a particular application design. [6] K. Hilgenberg, Sensor Data Collection Framework, 2019. [Online]. Available: https:
Another example is a multicriteria optimization of sensor systems //sourceforge.net/projects/sdcf/files/ProjectReport.pdf/download. Accessed: Sep.
9, 2019.
and networks. One of the traditional ways to improve the overall
[7] Z. Ma, Y. Qiao, B. Lee, and E. Fallon, “Experimental evaluation of mobile phone
data quality in SCADA and other data collection systems is fusing sensors,” in Proc. 24th IET Irish Signals Syst. Conf., Jun. 2013, pp. 1–8.
numerous data streams that originate from multiple networked sensor [8] Q. Mourcou, A. Fleury, C. Franco, F. Klopcic, and N. Vuillerme, “Performance
platforms. Various methods of sensor network optimization have been evaluation of smartphone inertial sensors measurement for range of motion,” Sensors,
proposed [24]–[26]. These methods could be improved if sensor qual- vol. 15, pp. 23168–23187, 2015.
[9] U. Shala and A. Rodriguez, “Indoor positioning using sensor-fusion in android de-
ity is evaluated, and these estimates are utilized in the sensor network vices,” master’s thesis, Dept. Comput. Sci., Kristianstad Univ., Kristianstad, Sweden,
design and optimization with the goal of improving the overall quality 2011.
of the collected sensor data [27]. [10] GSMArena.com—Mobile Phone Reviews, News, Specifications and More, 2019.
[Online]. Available: https://www.gsmarena.com/. Accessed: Nov. 30, 2019.
[11] Device Info HW Database, 2019. [Online]. Available: http://www.deviceinfohw.ru/
VII. CONCLUSION devices/index.php. Accessed: Nov. 30, 2019.
[12] Device Specifications—Mobile Device Specifications, Comparisons, News, User
Embedded sensors dramatically extend the functionality of mobile Reviews and Ratings, 2019. [Online]. Available: https://www.devicespecifications.
com/en. Accessed: Nov. 30, 2019.
devices and their market. The ubiquity of these sensors creates an [13] Catalogue for Smartphones—Specifications and Reviews—Phonesdata, 2019. [On-
excellent opportunity to develop various sensor networks with a wide line]. Available: http://phonesdata.com. Accessed: Nov. 30, 2019.
range of sensors employed by mobile devices’ users. Unfortunately, the [14] Tech News, Latest Technology, Mobiles, Laptops – NDTV Gadgets 360, 2019.
quality of these sensors affects their possible applications. In addition, [Online]. Available: https://gadgets.ndtv.com/. Accessed: Nov. 30, 2019.
[15] Technology News, Latest & Popular Gadgets Reviews, Specifications, Prices, Mobile
mobile device users may be unaware of sensors embedded in their
Comparison, Technology Videos & Photos | Gadgets Now, 2019. [Online]. Available:
devices and the quality of these sensors and, therefore, may choose https://www.gadgetsnow.com/. Accessed: Nov. 30, 2019.
the wrong sensors, which may lead to dangerous violations of their [16] Data Quality Lab—Sensor Dataset, 2020. [Online]. Available: https://drive.google.
security, privacy, and safety. In order to provide the sensor comparison com/drive/folders/12Sv03KZkzYJwoHlwNUFCtQAoBR7lbc1y?usp=sharing.
and recommendations on the sensor applications based on their quality, Accessed: Jan. 21, 2020.
[17] D. Nield, All the Sensors in Your Smartphone, and How They Work, 2019. [On-
we created and made available to the community the knowledge line]. Available: https://gizmodo.com/all-the-sensors-in-your-smartphone-and-
base of a wide range of sensors embedded in various mobile devices how-they-wor k-1797121002. Accessed: Nov. 30, 2019.
and their characteristics. This research facilitates closer collaboration [18] S. Davidian, How to Select a Humidity Sensor, 2019. [Online]. Available: http:
between sensor systems and network developers and sensor users, as //blog.servoflo.com/how-to-select-a-humidity-sensor. Accessed: Nov. 30, 2019.
[19] Applications and Uses of UV Sensors, 2019. [Online]. Available: https:
well as mobile device owners. The sensor analysis reveals that while
//www.apogeeinstruments.com/applications-and-uses-of-uv-sensors/. Accessed:
the quality characteristics of embedded sensors are lower than the Nov. 30, 2019.
specialized individually calibrated instruments in professional use, [20] Withings Pulse—What Does Spo2 Mean? What is a Normal Spo2 Level? 2019.
they may vary significantly. Based on the analysis, all sensors were [Online]. Available: https://support.withings.com/hc/en-us/articles/201494667-
classified into good and poor, and the integral quality indicator was Withings-Pulse-What-does-SpO2-mean-What-is-a-normal-SpO2-level-.
Accessed: Nov. 30, 2019.
calculated for each device. An Android application that analyzes the [21] P. Sadhukhan, “Performance analysis of clustering-based fingerprinting localization
quality of the sensors that are available on the device and provides this systems,” Wireless Netw., vol. 25, no. 5, pp. 2497–2510, 2019.
information to the device’s users has been developed. The developed [22] D. Arthur and S. Vassilvitskii, “k-means++: The advantages of careful seeding,” in
tool enables the education of sensor users and facilitates sensor’s secure Proc. 18th Annu. ACM-SIAM Symp. Discrete Algorithms, 2007, pp. 1027–1035.
[23] DataQualityLab, Sensor Quality Assessment—Apps on Google Play, Dec. 2019.
and safe applications.
[Online]. Available: https://play.google.com/store/apps/details?id=com.
dataqualitylab.sensorquality. Accessed: Dec. 9, 2019.
ACKNOWLEDGMENT [24] A. Bari, S. Wazed, A. Jaekel, and S. Bandyopadhyay, “A genetic algorithm based
approach for energy efficient routing in two-tiered sensor networks,” Ad Hoc Netw.,
This work was supported in part by the National Science Foundation under Award vol. 7, no. 4, pp. 665–676, 2009.
ACl-1547301. [25] N. Saxena, A. Roy, and J. Shin, “QuEST: A QoS-based energy efficient sensor
routing protocol,” Wireless Commun. Mobile Comput., vol. 9, no. 3, pp. 417–426,
2009.
REFERENCES [26] Y. Pan and X. Lu, “Energy-efficient lifetime maximization and sleeping scheduling
supporting data fusion and QoS in multi-sensornet,” Signal Process., vol. 87, no. 12,
[1] S. Olenski, The State of Crowdsourcing, 2015. [Online]. Available: https://www. pp. 2949–2964, 2007.
forbes.com/sites/steveolenski/2015/12/04/the-state-of-crowd sourcing/. Accessed [27] I. Khokhlov, A. Pudage, and L. Reznik, “Sensor selection optimization with genetic
on: Nov. 30, 2019. algorithms,” in Proc. IEEE Sens., Oct. 2019, pp. 1–4.

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 23,2020 at 21:18:27 UTC from IEEE Xplore. Restrictions apply.
View publication stats

You might also like