Sensing Gestures for Business Intelligence
David Bell, Nikhil Makwana and Chidozie Mgbemena
Department of Information Systems and Computing (DISC), Brunel University, London, U.K.
Keywords: Motion Sensors, Gestures, Business Intelligence.
Abstract: The combination of sensor data with analytic techniques is growing in popularity for both practitioners and
researchers as an Internet of Things (IoT) offers new opportunities and insights. Organisations are trying to
use sensor technologies to derive intelligence and gain a competitive edge in their industries. Obtaining data
from sensors might not pose too much of a problem, however subsequent utilisation in meeting an
organisation’s decision making can be more problematic. Understanding how sensor data analytics can be
undertaken is the first step to deriving business intelligence from front line retail environments. This paper
explores the use of the Microsoft Kinect sensor to provide intelligence by identifying and sensing gestures
to better understand customer behaviour in the retail space.
1 INTRODUCTION
Organisations generate vast amounts of data from
their day to day operations (McRobbie, et al., 2012;
Vera-Baquero, et al., 2013). From these data
sources, some organisations are able to harvest
insightful information termed Business Intelligence
(BI) (Feng & Liu, 2010). It has evolved into a set of
computer based techniques used to identify, extract
and analyse business data to better understand
current business trends and more importantly predict
future business patterns in order to gain competitive
advantages and effectively take steps to deliver
better performances. Globally dispersed customers
and extended supply chains have further motivated
the use of BI. Clearly, key data from customers
moving around the commercial environment is
missing when only focusing on completed
purchasing etc.
Importantly though, with a growing Internet of
Things (IoT) making our environment smarter, it
provides opportunities to analyse data that is
generated from smart objects such as sensors
(Doody & Shields, 2012). Increasingly, businesses
are discovering that environments and machines
augmented with sensors are able to send information
back to headquarters that once analysed can provide
a competitive advantage (Hewlett-Packard, 2013).
Making sense of the plethora of data generated by
such devices and leveraging reality mining
applications (Doody & Shields, 2012) can provide
an ability to extract useful knowledge from real
world sensor data and provide the basis to identify
predictable patterns. Sensor applications are widely
reported in literature with examples of location
technology and smartphone devices used to request
cabs, effectively saving cab drivers valuable time
and money in looking for fares (Hailo, 2013).
Within retail industries, footfall sensors are used to
provide intelligence for staff scheduling that ensure
customer demands are met without overheads being
excessive (Ipsos, 2013). Together with an overall
trend in Point-of-Sale (POS) transactions, loyalty
cards used by major retailers are being utilised to
track customer purchases leading to a greater
understanding of shopping habits and enabling
businesses to track the success of marketing
initiatives (Zakaria, et al., 2012).
Despite uncovering a wealth of data about
shopper demographics and their purchase history,
insight into the customers in store behaviour is often
missing. Understanding where customers spend
most of the time when within a store, identifying
which products and promotional displays are most
popular, how long customers wait in lines etc. can
address this lack of ‘front-line’ knowledge. This
paper explores how motion sensors be used to better
understand the retail environment. Leveraging
motion sensing technology provides the ability to
precisely track customers and observe human
behaviour for pattern analysis and business
intelligence (Cao, 2008). The paper reports on
related work in both motion sensing and business
intelligence before coverage of our specific pattern
52
Bell D., Makwana N. and Mgbemena C..
Sensing Gestures for Business Intelligence.
DOI: 10.5220/0004878100520060
In Proceedings of the 3rd International Conference on Sensor Networks (SENSORNETS-2014), pages 52-60
ISBN: 978-989-758-001-7
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
sensing approach and framework. A practical
experiment is then presented that leads to our BI
architecture discussion and concluding remarks.
2 RELATED WORK
Two areas of interest in this research paper are
Business Intelligence (BI) and Motion Sensors
technology. In light of this, the purpose of this
section is to highlight business intelligence currently
used by businesses and identify architectures used in
case studies. Secondly, research on motion sensors is
used to identify suitable sensors that are able to
provide motion data to a business intelligence
framework.
2.1 Business Intelligence
Luhn coined the term ‘Business Intelligence’ (BI) in
1958 (Luhn, 1958). He defined BI as “the ability to
apprehend the interrelationships of presented facts in
such a way as to guide actions towards desired
goals” (Luhn, 1958, pp. 314-319). Despite having
less significance at the time (when compared to its
relevance today), this key work underpins the theory
and practice of decision support systems. The work
carried out by Luhn has been fundamental in
providing reporting, Online Analytical Processing
(OLAP) and visualisation capabilities used business
decision making today. As a modern day term, BI is
used to describe the collection of processes, tools
and technology helping in achieving more profit
through improving productivity, sales and services
of an enterprise. With the aid of BI methods,
corporate data can be better analysed and
transformed into useful knowledge, which in turn is
required to achieve a profitable business action
(Martin, et al., 2012). Modern day business
intelligence tools include applications such as
Executive Information Systems (EIS), Customer
Relationship Management (CRM) and Corporate
Performance Management (CPM) (Tvrdikova,
2007). What all of these commercial applications
have in common is the ability to use multi-
dimensional data stores that have the benefits of
reporting rapid changes as opposed to waiting on
batch jobs that run overnight. However, despite their
rapid response to change, drawbacks include
increased complexity (Vassiliadis, 1998) and
maintenance.
Business Intelligence systems are often regarded
as business critical solutions and invaluable tools,
adding significant value to an enterprise. These
systems have often become of paramount
importance and are now considered a vital tool for
day-to-day business (Xu, et al., 2007). The main
purposes of employing business intelligence tools
are to provide the right information to the right
people at the right time, with the overall aim of
allowing effective decision making to take place
(Xu, et al., 2007; Krneta, et al., 2008). Many
examples of successful BI implementation include
Netflix (an on-demand internet streaming media
provider) that leverage analytics to analyse customer
behaviour to better serve its subscriber base by
recommending the right options for their viewers
(Microstrategy, 2013). Carphone Warehouse (an
independent retail telecommunications provider)
utilise business intelligence reporting tools for
making decisions such as adjusting headcount based
on footfall traffic and conducting in store employee
reviews for tracking performance (Computing,
2012). Barclays Bank has employed the use of BI
systems to reduce operational costs helping them in
efficient recruitment planning and training schedules
(Ridgian, 2011).
BI is a necessity in the retail industry. It is an
essential element when realising typical in store
activities such as those that further identify potential
intelligence that could be used to provide
competitiveness. Burke (2002) identifies consumer
behaviours such as: 1) Entering the store, 2) entering
a specific aisle, 3) checking out and paying for
items, and 4) post purchase of customer service
(Burke, 2002). These behaviours highlighted by
Burke (2002) are also applicable to
telecommunication retail environments however
consumers are viewing either live or dummy phones
(PhoneDog Media, 2013).
Many organisations now rely on and are heavy
users of Business Intelligence systems. Tesco for
example, is a multinational supermarket and now
rated as the second most powerful retailer on the
planet. A primary factor in their success is having BI
systems that enable them to gather data and
analytical skills to test ideas and turn insights into
customer and business relevant actions (Advanced
Performance Institute, 2013).
Leonidas, et al. (2012) present a middleware
framework named ‘PLATO’ to provide business
intelligence applications. “The main goal is to
provide a framework for the development of
applications involving heterogeneous sensors”
(Leonidas, et al., 2012, pp. 266-270). The
framework supports collection, processing and
analysis of the data originating from sensors while
enabling the development of business intelligence.
SensingGesturesforBusinessIntelligence
53
Wang, et al., (2012) present a system that obtains the
characteristic travel behaviour of an urban
population and builds a framework using large scale
transportation datasets to extract value added
information with the ultimate aim of reducing fuel
consumption, improving customer satisfaction and
business performance. Hondori et al., 2012 present a
tele-rehabilitation study that monitors post stroke
patients using sensor technology to provide
intelligence in monitoring their activities of daily
living and progress in a home environment. Despite
many studies presenting research from post stroke
patients using wearable sensors, (O’Keeffe, et al.,
2007; Zhang, et al., 2012; Hester, et al., 2006), there
are many reported drawbacks including the wearable
sensor being obtrusive, heavy, costly and generally
inconvenient (Benning, et al., 2007; Hondori, et al.,
2012). Hondori, et al., 2012 has leveraged motion
sensor technology using a Kinect sensor which is
completely wireless and a non-wearable technology
that provides a more natural experience to sense the
health status of the patient (Hondori, et al., 2012).
2.2 Motion Sensing
“Motion detection is the fundamental process of
detecting whether any entities exist and are moving
around the area of interest” (Xiao, et al., pp. 229-
235). The motion sensor/detector is a converter that
measures a physical quantity and coverts it into a
signal form which can be read by an electronic
instrument.
Various sensors exist for different purposes such
as accelerometers, gyroscopes and magnetometers.
Accelerometers are a type of device that measures
the acceleration. The modern day use of
accelerometers can be seen in user interface control
on smartphone technology as the tilting motion is
used to differentiate between portrait and landscape.
A gyroscope can be used to either measure, or
maintain, the orientation of a device. Unlike an
accelerometer, which measures the linear
acceleration of the device, a gyroscope measures the
orientation directly. Gyroscope technology has
created the ability of making the gaming experience
as real as possible. Magnetometers measure the
ambient magnetic field and provide digital compass
like applications on smartphone devices. These are
also more traditional sensors which are small and
come at a reasonable cost (Amma, et al., 2010) –
when compared to smartphones or tablets. The first
generation of sensors included pen based and hand
worn sensors known as ‘data gloves.’ However, the
latter form of these sensors has limitations such as
the feeling of unnaturalness, obtrusive and generally
a burden to wear (Ren, et al., 2013). In contrast to
the data gloves, and with the recent advancements in
HCI and motion sensor technology, many products
have surfaced, repurposing traditional sensing
technology into new forms. This is more prevalent in
gaming systems such as the Nintendo Wii Remote,
PlayStation Eye/Move and the Microsoft Kinect for
Xbox; each offering a more natural experience and
touch free way of communication.
The Microsoft Kinect sensor, released in 2010
for the Xbox, enables users to control and interact
with the Xbox 360 without the need for any physical
interaction of game controllers. This is realised by
the use of a natural user interface and using gestures
and spoken commands (Microsoft Corporation,
2013). In contrast, the Wiimote and PS Move
require physical contact for navigation. In 2011
Microsoft released a non-commercial software
development kit (SDK) for the Kinect sensor
allowing developers to build natural, intuitive
computing experiences using C#, C++ and Visual
Basic programming languages.
3 SENSING PATTERNS
This research project employs an objective-centred
approach to the Design Science Research
Methodology (DSRM) as outlined by Peffers et al.
(2007). The design problem is identified from the
literature review and in this case includes a lack of
smart analysis when providing business intelligence
in a ‘front-line’ retail space. Traditional technology
is still being used with an over-reliance on EPOS
systems and data. Our design explores how recent
gaming sensors can provide more effective sensory
input – with a specific focus on people in the retail
environment.
3.1 Sensing Architecture
A high level architectural overview is presented in
Figure 1. The framework separates the collection of
sensor data from a gesture analysis and codification
phase. A sensor augmented environment is able to
provide a basis for gesture identification rules. Once
identified, the gestures can be measured and
reported by a BI system.
The user is free to use their natural interactions
in the retail environment which includes activities
such as viewing in-store adverts, paying for products
at a cashier till and extracting dummy phones for
closer viewing. The Kinect sensor (from Microsoft)
SENSORNETS2014-InternationalConferenceonSensorNetworks
54
Figure 1: High Level Architecture.
captures these movements/gestures with the help of
the Kinect SDK. The depth stream available to the
Kinect sensor is used to process the depth data to
generate skeleton data of human body joints, which
is stored as X, Y and Z values using the C#
application created into a text file. Offline analysis
of the raw co-ordinate dataset is analysed (with the
use of R) to highlight rules that can automatically
identify user gestures. Observed analysis results are
then written into the sensing application to identify
only the rules identified from the analysis in order to
provide meaningful business intelligence of
customer activity.
3.2 Sensor Data Capture
In order to capture a customer’s natural interaction
in a retail space, a Kinect application has been
developed as shown in Figure 2. The user interface
has been kept relatively simple as this design phase
is focusing on the framework processes required to
undertake this new form of BI analysis. The main
functionalities of the application:
Provide 3 streams (RGB, Skeleton and Depth) to
provide visual feedback to the user for calibration
purposes.
Seated mode option to capture the top half of
skeleton (10 joints)
Kinect elevation angle to also provide in
calibration purposes.
Capture button, to provide the saving of tracked
skeleton body co-ordinates to a data file.
The Kinect sensor provides an easy to use API
and ready access to co-ordinates for each part of the
body, e.g. elbow_right.Position.X.
Initially, these co-ordinates are collected and stored
in a data file for rule identification.
Figure 2: Sensor Data Capture Application.
3.3 From Data to Gestures
The popular R software environment is typically
used for statistical computing and graphics. R has
been used in this research project for statistical
analysis and primarily visualisation of the body
coordinates that have been captured from Kinect
sensors in the retail environment. Command line
scripting is used for the analysis of sensor data and
rule identification. A third party package ‘ggplot2’
has been utilised in R to provide graph building
capabilities.
Although we have built a system with a single
sensor, it is envisaged that the same approach will
allow for multiple sensors in many retail
environments. Importantly, an IoT that connects
many motion sensors requires an architecture that is
able to identify gestures before forwarding to
integrated reporting platforms.
The resulting visualisations of sensor data allow
the programmer or data analyst to identify rules for
specific customer actions (gestured)– e.g. Browsing
and reaching for a phone. The graphical output can
be seen in Figure 3 and includes joints co-ordinates
in different colours. Rule identified in this process
SensingGesturesforBusinessIntelligence
55
Table 1: R Code Descriptions.
Browsing a Phone Display
ggplot(data=mydata,
aes(x=Time)) +
geom_line(aes(y=Hand.Rig
ht.Y,
colour="Hand.Right.Y",
group=1)) +
geom_line(aes(y =
Hand.Left.Y,
colour="Hand.Left.Y",
group=1)) +
geom_line(aes(y =
Head.Y, colour="Head.Y",
group=1)) +
geom_line(aes(y =
Spine.Y,
colour="Spine.Y",
group=1)) +
ylab("Coordinates") +
opts(title="Browsing
Demo Phones
Coordinates")
Time variable has
been used as the X
axis of the graph
against the other
joints listed in the
command line as the
Y axis. The
commencing
command displays
the colour of the
plot line along with
the variable and its
position.
are then coded within the same application presented
in Figure 2, adding the recorded gesture to the
output data file. Once coded, the application is now
ready to capture gestures for use in BI analysis and
reporting. The Kinect sensor and application can
now be started in the retail space.
Figure 3: Sensor Data Visualisation.
4 EXPERIMENATION
Instantiations operationalize constructs, models and
methods (March & Smith, 1995) – typically creating
a working implementation. Here we can use the
implementation to evaluate the effectiveness of the
design – a new sensor driven BI framework.
4.1 Retail Business Context
The aim of the lab experimentation was to define
gestures in terms of skeletal joint movement through
subject’s natural interactions via the use of motion
capture. Encompassed within this is the
identification of gesture rules (e.g. the heuristics for
gesture identification). The experiment took place
in the lab environment (see Figure 4). The aim of the
lab environment is to simulate a real world shopping
(Mobile Telephone Shop) experience for the
purposes of capturing interactions. A whiteboard has
been used to simulate a phone wall display, Point of
Sale (PoS) cashier and a gaming experience. A wall
has been used to display phone advertisements for
subjects to browse. A Kinect sensor has then been
mounted on top of a PC monitor, which provides
motion sensing capacities that capture the subject
standing in the simulated store environment. The
sensed user will have their skeleton joints captured
for motion activity and will be saved in a text format
as it allows interoperability with other applications.
Figure 4: Simulated Retail Environment.
4.2 Captured Data
A number of scenarios were tested in the simulated
environment in order to: 1) Capturing data with the
application, 2) analyse the data in order to identify
rules, 3) code the rules in the applications and then
4) capture gestures. The scenarios (and rules) can be
seen in Table 2. Importantly though, it is the
observation of humans early in this process that
allow for systematic analysis of key activities – and
their associated gestures.
The various joints of the body are analysed to
uncover rules for scenario actions. These are then
coded within the capture application.
SENSORNETS2014-InternationalConferenceonSensorNetworks
56
Table 2: Scenarios and rules identified.
Scenario
Rule
Viewing In
store Ads
Either hand extending out for small
constant period.
Hand extending outwards at the same
position as the head.
Playing
Console
Games
Head and spine joint stay roughly in the
same position throughout activity.
Sudden spikes of hands going up indicate
a process of celebration as well as
expression of sadness when losing in
game.
Returning
Product at
Cashier
The right hand is observed to move in
front of spine and upwards towards the
counter for a moment (returning the
product).
The right hand is observed again to move
in front of spine and upwards towards the
counter which identifies the payment
process.
Paying for
Product at
Cashier
The right hand is observed to be extracted
near to both hips possibly indicating
reaching for wallet.
Right hand is then seen in front of spine
which suggests a payment is in progress
(handing card over to assistant) this is
also indicated by a period of stability as
the right hand is raised indicating a
payment.
Browsing
Demo
Phones
Left/Right hand is shown progressing
away considerably far from the head
(reaching and extracting demo phones
from display).
Right hand is then settled immediately in
front of head as it is being viewed.
Discussion
on Shop floor
Left/Right hand shows movement that is
directly beside of spine. Right hip seen
mostly in one place, also replicated
behaviour in the spine.
Right hand is extracted down next to right
hip.
The resulting system, with rules encoded, is able to
then capture that data necessary for BI reporting.
Iterating through each scenario resulted in a
framework for sensor data analysis and codification
(presented in Table 3).
Figure 5: Sensor Visualisation for rule identification.
Table 3: Sensor BI Framework.
Process Description
Platform
Acquisition
A motion capture solution is created or
acquired to support the capture of
customer’s natural interaction by
tracking their skeleton movement.
Skeletal
Capture
A motion sensor placed in the retail
environment captures customer’s
natural form of interaction. The capture
produces full bodily coordinates that
will be stored for post capture analysis.
Post-
capture
Human
Analysis
Joint coordinates are visualised in a
graph form in order to identify trending
patterns on the different bodily skeletal
joints. The trends provide a means to
define gesture rule identification.
Rule
Codification
The rules identified from the post
capture analysis are implemented into
the capture application, enabling the
capture to identify gestures triggered
by subject’s natural interaction.
Recapture
A motion sensor is placed in the retail
environment that captures gestures
(based on the earlier rule
identification).
End of Day
(EOD)
Business
Intelligence
Reporting
Aggregated gestures identification
reports are produced for viewing by
managers/stakeholders (or integration
with other data sources within a BI
tool)
4.3 Business Intelligence
A number of visualisations were developed for
illustrative purposes. Primarily to put the data
capture process in a business context. A sample of
these is shown in Figure 6. The textual output was
easily loaded into commercial grade BI reporting
SensingGesturesforBusinessIntelligence
57
products (which is our case was Tableau).
Figure 6: BI Reporting.
Experimenting with our business intelligence views
and using a new group of human subjects, a bar
chart depicting hourly recognised gestures has been
created to identify the success rate of gesture
recognition (%) throughout the day. The average
identification success rate is just over 50%.
Although disappointing, this was largely due to the
Kinect application calibrated for the few subjects
that took part in the initial test. Despite this, the
framework has proved effective ‘in use’ and offers
interesting possibilities when part of a wider
business intelligence environment, linking data with
further data sources.
4.4 Discussion and Evaluation
The paper has made a contribution in the research
area of motion sensing in a retail space environment.
Utilising the Natural User Interface (NUI) paradigm
commonly found in the gaming industry and
applying it in a retail setting is an effective means of
gathering business intelligence. This paper has
identified that commodity sensor hardware, which in
this case is the Kinect motion sensor, offers a novel
approach to providing strategic organisational
insights into customer activity in the field. This has
the potential to offer substantial competitive benefits
are more nuanced customer interactions are visible.
Production of business intelligence reports provides
additional assistance to the traditional POS reporting
methods currently relied upon and promotes novel
intelligence gathering perspectives. Resulting from
the research is a novel framework that uncovers and
reports on gesture based business intelligence – also
having a wider applicability. The research
contributions of this research paper can be easily
extended and applied practically in other domains,
for example the health sectors and specifically tele-
health for purposes such as monitoring of post-
stroke patients.
4.5 Future Research
There is scope for this research to be progressed
further to form part of the larger business
intelligence platform. During the development
phase, areas of further work were identified:
1) Social Media – The rise of ‘citizen sensor
networks’ provides an opportunity to
understand and analyse data reported by citizen
sensors and the fusion of this data with the
gesture sensed data to identify further potential
trends. Gathering intelligence in this manner
may be able to add a new perspective,
identifying novel business intelligence
(combining physical action and opinion).
2) Data repositories – With the aforementioned
fusion of social data, data repositories stored by
organisations such as transaction histories,
customer data, and internal ERP systems can
also be integrated and fused into the sensed data
which then gives the possibility of building
customer profiles from past data. Data gathered
can be used by many departments in for-profit
organisations, such as marketing departments
for effective use of advertising.
3) Multiple sensors – Employing the use multiple
Kinect (or other) sensors and fusing the
captured date from the sensors is highly likely
to enhance the accuracy of the results obtained.
5 CONCLUSIONS
This paper presents an approach for monitoring and
understanding customer behaviour in-store using
motion sensors. In the design experiment presented,
users performed natural activities in a retail space
that included actions such as browsing phones,
viewing ads and paying for products at the cashier
till. The experiment was carried out in a lab
environment with posters of phones, ads and cashier
point to simulate a retail space. A Kinect sensor was
used to capture movements and R was used to define
rules which automatically captured user gestures. A
framework is outlined that separates the capture of
motion from the gesture identification and business
intelligence reporting. A number of interesting
avenues of further research are apparent from our
SENSORNETS2014-InternationalConferenceonSensorNetworks
58
exploratory work – adding intelligent data analysis
to automatically identify rules as well as integrating
related data sets.
REFERENCES
Advanced Performance Institute, 2013. Business Intelligence
(BI) - What is BI? Training, examples & case studies:
[Online] Available at: http://www.ap-institute.com/
Business Intelligence.html (Accessed 25 July 2013).
Amma, C., Gehrig, D. & Schultz, T., 2010. Airwriting
Recognition using Wearable Motion Sensors. ACM, pp.
10-11.
Benning, M. et al., 2007. A Comparative Study on Wearable
Sensors for Signal Processing on the North Indian
Tabla. Communications, Computers and Signal
Processing.
Burke, R. R., 2002. Technology and the customer interface:
what consumers want in the physical and virtual store.
Journal of the Academy of Marketing Science.
Cao, L., 2008. Behavior Informatics and Analytics: Let
Behavior Talk. Data Mining Workshops.
Computing, 2012. Case Study: Business intelligence at
Carphone Warehouse - 23 Aug 2012 - Computing
News:. (Online) Available at: http://
www.computing.co.uk/ctg/news/2200696/case-study-
business-intelligence-at-carphone-warehouse (Accessed
02 September 2013).
Doody, P. & Shields, A., 2012. Mining Network
Relationships in the Internet of Things. ACM.
Feng, Y. & Liu, Y., 2010. Design of the Low-Cost Business
Intelligence System Based on Multi-agent. Information
Science and Management Engineering (ISME).
Hailo, 2013. HAILO. The Black Cab App:. (Online)
Available at: https://www.hailocab.com/
(Accessed 1 September 2013).
Hester, T. et al., 2006. Using Wearable Sensors to Measure
Motor Abilities following Stroke. Wearable and
Implantable Body Sensor Networks.
Hewlett-Packard, 2013. From the Internet of Things, a
business intelligence bounty – HP Software Discover
Performance:. (Online) Available at: http://
h30458.www3.hp.com/us/us/discover-performance/info-
management-leaders/2013/apr/from-the-internet-of-
things--a-business-intelligence-bounty.html (Accessed 1
September 2013).
Hondori, H. M., Khademi, M. & Lopes, C. V., 2012.
Monitoring Intake Gestures using Sensor Fusion
(Microsoft Kinect and Inertial Sensors) for Smart Home
Tele-Rehab Setting. IEEE.
Ipsos, 2013. Why count customers? | Experts in retail footfall
counting | IPSOS Retail Performance:. (Online)
Available at: http://www.ipsos-retailperformance.com/
WhatWeDo/WhyCountCustomers (Accessed 01
September 2013).
Krneta, D., Radosav, D. & Radulovic, B., 2008. Realization
business intelligence in commerce using Microsoft
Business Intelligence. Intelligent Systems and
Informatics, pp. 1-6.
Leonidas, P., Drakoulis, D., Dres, D. & Smailis, C., 2012.
PLATO - Intelligent Middleware Platform for the
Collection, Analysis, Processing of Data from Multiple
Heterogeneous Sensor Systems and Application
Development for Business Intelligence. Informatics
(PCI).
Luhn, H. P., 1958. A Business Intelligence System. IBM
Journal of Research and Development, Volume 2, pp.
314-319.
March, S. T. & Smith, G. F., 1995. Design and natural
science research on information technology. Decis.
Support Syst., Volume 15, pp. 251-266.
Martin, A., Lakshmi, M. & Prasanna Venkatesan, V., 2012.
An analysis on business intelligence models to improve
business performance. Advances in Engineering,
Science and Management (ICAESM), pp. 503-508.
McRobbie, G., Talati, S. & Watt, K., 2012. Developing
business intelligence for Small and Medium Sized
Enterprises using mobile technology. Information
Society (i-Society).
Microsoft Corporation, 2013. Product Features | Microsoft
Kinect for Windows:. (Online)
Available at: http://www.microsoft.com/en-
us/kinectforwindows/develop/sdk-eula.aspx (Accessed
20 July 2013).
Microstrategy, 2013. MicroStrategy-Mobile-BI-Retail-
Apps.pdf:. (Online) Available at: https://
www.microstrategy.com/Strategy/media/downloads/pr
oducts/MicroStrategy-Mobile-BI-Retail-Apps.pdf.
O’Keeffe, D. T., Gates, D. H. & Bonato, P., 2007. A
Wearable Pelvic Sensor Design for Drop Foot
Treatment in Post-Stroke Patients. Engineering in
Medicine and Biology Society.
PhoneDog Media, 2013. How important are display models
in wireless retail stores? | PhoneDog:. (Online)
Available at: http://www.phonedog.com/2013/
04/23/how-important-are-display-models-in-wireless-
retail-stores/ (Accessed 02 September 2013).
Peffers, K., Tuunanen, T., Rothenberger, M. A. & Chatterjee,
S., 2007. A Design Science Research Methodology for
Information Systems Research. J. Manage. Inf. Syst.,
Volume 24, pp. 45-77.
Ren, Z., Yuan, J., Meng, J. & Zhang, Z., 2013. Robust Part-
Based Hand Gesture Recognition Using Kinect Sensor.
Multimedia, IEEE Transactions on, Volume 15, pp.
1110-1120.
Ridgian, 2011. RidgianBarclaysYourPeopleCS.pdf:. (Online)
Available at: http://www.ridgian.co.uk/downloads/
RidgianBarclaysYourPeopleCS.pdf (Accessed 03
September 2013).
Tvrdikova, M., 2007. Support of Decision Making by
Business Intelligence Tools. Computer Information
Systems and Industrial Management Applications, 2007.
Vassiliadis, P., 1998. Modeling multidimensional databases,
cubes and cube operations. Scientific and Statistical
Database Management.
Vera-Baquero, A., Colomo-Palacios, R. & Molloy, O., 2013.
Business process analytics using a big data approach. IT
Professional.
SensingGesturesforBusinessIntelligence
59
Wang, C. & Chen, H., 2012. From data to knowledge to
action: A taxi business intelligence system. s.l.,
Information Fusion (FUSION).
Xiao, J. et al., 2012. FIMD: Fine-grained Device-free
Motion Detection. Parallel and Distributed Systems
(ICPADS), pp. 229-235.
Xu, L. et al., 2007. Research on Business Intelligence in
enterprise computing environment. Systems, Man and
Cybernetics, pp. 3270-3275.
Zakaria, I., Rahman, B. A. & Othman, A. K., 2012. The
relationship between loyalty program and customer
loyalty in retail industry: A case study. Innovation
Management and Technology Research (ICIMTR).
Zhang, Z., 2012. Microsoft Kinect Sensor and Its Effect.
MultiMedia, IEEE, pp. 4-10.
SENSORNETS2014-InternationalConferenceonSensorNetworks
60
