Information Technology Reference
In-Depth Information
3.3.2 Sensor Type and Smartphones
There is currently awide range of sensors available that have been used for developing
HAR systems. In this review, we highlight the ones exploited for wearable sensing.
Inside this category, we find that sensors can come as a self-contained device, or take
part of specific purpose hardware along with other sensors (e.g. (Sama et al. 2012 )),
or be interconnected as nodes forming a Body Sensor Network (BSN) (Yang and
Yacoub 2006 ). Moreover, they can also be embedded elements within other portable
devices, such as smartphones or Personal Digital Assistants (PDAs). In this section,
we contrast the various wearable approaches used for gathering activity-related sig-
nals. Previous HAR works selected accelerometers as their preferred inertial sensor,
e.g. (Allen et al. 2006 ;Ravietal. 2005 ; Sama et al. 2012 ). However, they are gen-
erally employed in cooperation with other sensors such as gyroscopes, microphones
and vital signs sensors, in order to contribute with additional information in the
recognition process (Lara et al. 2012 ; Lee and Mase 2002 ; Lukowicz et al. 2004 ).
Previous work has used configurations of multiple sensors located in different
body parts, generally ranging from1 to 5. Although the use of numerous sensors could
improve the performance of a recognition algorithm, it is unrealistic to expect that the
general public will use them on a regular basis as their obtrusiveness is relatively high
and it becomes tedious to wear them. For instance, in Bao and Intille ( 2004 ), where
a set of five biaxial body-worn accelerometers was used, a drawback was evident
regarding the number of sensors they needed to attach around the body. The same
occurs with other works such as (Lukowicz et al. 2004 ; Mannini and Sabatini 2010 ;
Salarian et al. 2007 ). The trend shows that in recent years, HARapproaches are aiming
to reduce obtrusiveness, either by using less wearable sensors, by opportunistically
gathering signals from commonly used devices (e.g. smartphones) or by even fusing
sparse sensor data from environmental sensors when available (Bahle et al. 2013 ).
In our approaches (e.g. L-HAR, PTA-HAR), we only use a single device with only
two embedded inertial sensors (accelerometer and gyroscope) and explore how it is
possible to achieve HAR with them.
Research efforts have also concentrated on exploiting smartphones for HAR.
Some smartphone-based approaches have been already proposed in the literature
(Berchtold et al. 2010 ; Brezmes et al. 2009 ; Kwapisz et al. 2011 ;Wuetal. 2012 ).
In Kwapisz et al. ( 2011 ), for example, it was presented one of the first approaches
to exploit an Android smartphone and its embedded triaxial accelerometer for HAR.
Their approach was able to classify six locomotion activities over intervals of 10 s
using an ANN while the device was carried in the pockets. In Nham et al. ( 2008 ),
the mode of transport (walking, running, cycling and driving) was predicted using
accelerometer data from an iPhone through a signal Fourier analysis and an SVM
classifier. In a similar way, Brezmes et al. in ( 2009 ) implemented a real-time activity
classifier to detect 6 different states on a Nokia mobile phone. Research on HAR
with smartphone has mostly incorporated only accelerometers (Lara and Labrador
2012a ; Mannini and Sabatini 2010 ). This can be explained due to the fact that these
embedded inertial sensors were the first to be introduced in the mobile phone market
 
Search WWH ::




Custom Search