IEEE Signal Processing - May 2018 - 78

devices and has been used for indoor event detection [11],
[53]-[58]. However, most aforementioned CSI-based indoor
sensing systems rely on only the CSI amplitudes, whereas the
phase information is discarded regardless of how informative
it may be.
Another technology category of device-free indoor monitoring systems is adopted from radar imaging technology
to track targets [59], [60], [70], [71]. However, their techniques consume over 1 GHz of bandwidth and require a
specially designed RF signal to sense the environment.
Recently, Ohara et al. introduced a new method for indoor
event detection by recognizing and classifying CSI with
a deep neural network (DNN) composed of three convolutional layers and two recurrent layers [72]. However,
because of the nature of DNNs, the complexity of training data collection and network learning is extremely high.
In contrast, by leveraging the TR technique to exploit rich
CSI, the proposed TRIEDS introduces a novel and practical
solution that can well support through-the-wall detection
and requires only low-complexity, single-antenna hardware
operating in the ISM band.

Human radio biometrics
Traditional biometrics systems, including fingerprint, face,
and iris recognition, are accurate. However, they require special devices to capture human biometric traits in an extremely
LOS environment, i.e., the subject must make contact with the
devices. Another category of biometrics is gait analysis. It
relies on the individual walking pattern to distinguish identity.
Conventional gait recognition requires high-speed cameras,
wearable sensors, and floor sensors [73].
Recently, gait recognition has been extended to an RF platform where the Doppler shift or the ToF of the signal reflected
by the human body is used to extract the individual gait pattern
[70], [74], [75]. However, to get a high-resolution gait profile, it
relies on special devices to scan over an ultrawide spectrum, and
LOS transmission is often required to guarantee the accuracy of
the extraction. Moreover, the computational complexity introduced by the necessary image processing and machine learning
algorithms for gait recognition is high. Unlike the aforementioned system, the proposed human identification system is
based on radio biometrics and is thus capable of distinguishing
and identifying different individuals through walls accurately
with commercial Wi-Fi devices. In addition, the proposed system can support simple and efficient algorithms to achieve highaccuracy performance.

Breathing monitoring
Contact-free breathing monitoring schemes have been developed to overcome the drawbacks of conventional breathing
monitoring methods requiring physical contact with human
bodies. Among them, schemes using EM waves are favorable, since they can monitor breathing rates through the
wall in a highly complicated indoor environment. In terms
of techniques, they can be classified into radar based and
Wi-Fi based.
78

Among the radar-based schemes, Doppler radar is commonly used, which measures the frequency shift of the signals
caused by the periodic variations of the EM waves reflected
from human bodies [77]. Recently, Adib et al. presented a
vital signs monitoring system that uses the Universal Software
Radio Peripheral as the RF front end to emulate a frequency
modulated continuous radar [78]. However, the requirement of
specialized hardware introduces a significant deployment cost.
In addition, the regulation on the transmission power significantly limits the range the system can monitor.
For the Wi-Fi-based schemes, RSSI is often used be cause of its availability on many Wi-Fi devices. In [79], Ab delnasser et al. presented UbiBreathe, which harnesses
RSSI on Wi-Fi devices for breathing estimation. However,
UbiBreathe is accurate only when users hold the Wi-Fi
device close to their chest. More recently, CSIs were used
for breathing monitoring. The scheme proposed by Liu et al.
in [80] is one of the first few CSI-based breathing monitoring
approaches. Nevertheless, a periodogram is used for spectral analysis, which needs a relatively long time for accurate
breathing monitoring.
Compared with the aforementioned methods, the proposed
breathing monitoring scheme is infrastructure free, since it
utilizes off-the-shelf Wi-Fi devices. Furthermore, with the
Root-MUSIC algorithm, the proposed approach can achieve
highly accurate breathing rate estimations within a short period of time, and it can resolve the breathing rates of multiple
people concurrently.

Conclusions
As a revolutionary platform that connects everything around the
world, the IoT has dramatically changed our lifestyle and
enabled us to measure and track everything connected to it.
Because of the ubiquitous deployment of wireless devices, wireless sensing that can make many smart IoT applications possible has recently received a great deal of attention. As the next
generation of wireless systems embraces a larger bandwidth, richer information can be revealed through wireless sensing, e.g., in
the form of multipaths. As bandwidth increases, the number of
multipaths that can be resolved also increases, allowing them to
serve as hundreds of virtual antennas.
Motivated by the physical principle of TR, we developed
various radio analytics for smart IoT applications in indoor
positioning, event detection, human recognition, and vital
signs monitoring. Unlike conventional approaches for these
applications, the proposed radio analytics approach can work well
under NLOS and enjoys low implementation complexity, thus
making it an ideal paradigm for smart IoT sensing, positioning,
and tracking.

Authors
Beibei Wang (bebewang@umd.edu) received her B.S. (highest
honors) degree in electrical engineering from the University of
Science and Technology of China, Hefei, in 2004 and her Ph.D.
degree in electrical engineering from the University of
Maryland, College Park, in 2009. She was a research associate

IEEE Signal Processing Magazine

May 2018

Table of Contents for the Digital Edition of IEEE Signal Processing - May 2018