IEEE Signal Processing - May 2018 - 62

impulse) to transceiver A, at which the multipath CSI [44] can
be estimated. Then transceiver A time-reverses the received
waveform (and conjugates it, if the signal is complex) and
transmits the time-reversed version of the waveform back to
transceiver B. The work in [40] and [42] has shown that the
convolution of the time-reversed waveform and the channel
can generate a unique peak at the specific receiver's location,
called the spatial focusing effect. This indicates that the multipath channel profile works as a unique and location-specific
signature and that the spatial focusing effect happens only
when the channel can match the time-reversed waveform. By
comparing the multipath CSI with a set of time-reversed CSI
values precollected at multiple known locations, one can infer
the current location of a device, and this idea can be applied to
assist positioning.
Since each multipath profile is, in essence, a focusing point
on the TR logical space, if there is an event that affects the multipath, such as a door opening or closing, the multipath profile
becomes mapped to another focusing point. If one can perform analytics or machine learning to distinguish both events,
then one is able to infer what has happened. With this notion,
we can further design various types of radio analytics based
on the multipath CSI. By fully exploiting the rich multipath
CSI, this approach can decipher the propagation environment
as if it were a sixth human sense, revealing subtle information
on various human activities. Thus, radio analytics can enable
many cutting-edge IoT applications, such as accurate indoor
positioning, tracking, wireless event detection, human recognition, and vital signs monitoring, as we will illustrate.

accuracy, even with a single AP working in the NLOS condition. However, TRIPS implementation requires a large bandwidth and specialized hardware. One would naturally ask if we
can simply use off-the-shelf Wi-Fi to accomplish this. In this
section, therefore, we discuss a centimeter-accuracy IPS using
commercial Wi-Fi, where we utilize a radio analytics based
on a combination of the CSI values from different frequency
bands or antenna links.

Impact of effective bandwidth
As discussed earlier, a large bandwidth is indispensable for
resolving multipath CSI and high correlation of CSI values
from different locations, because insufficient bandwidth can
result in positioning ambiguity. To understand how bandwidth
affects the accuracy of indoor positioning, we conducted
extensive experiments in a typical indoor space. We deployed
two channel sounders under an NLOS setting, with one of
them placed on an experimental structure with a 5-mm measurement resolution. At each location of the experimental
structure, we collected multiple CSI values as location-specific fingerprints under different bandwidths. To assess the similarities among locations, we defined an analytic value of the
CSI values belonging to two locations-the TR resonating
strength (TRRS) [31]-as
TR (h F, 1, h F, 2) = c

Today, mobile devices have become an indispensable part of
daily life. To provide users with seamless services through
such gadgets, it is crucial for service providers to know an
individual's exact location. However, locating a person in an
indoor environment is quite challenging, because signals are
attenuated by the numerous multipaths in the rich-scattering
indoor environment.
Most existing IPSs can achieve only meter-level accuracy,
and performance becomes even worse in the non-LOS (NLOS)
condition. The main reason is that it is generally difficult or even
impossible to obtain precise measurements because of the richscattering indoor environment, and such imprecise measurements lead to ambiguity when running positioning algorithms.
To reduce ambiguity, one can collect more online measurements
and/or deploy multiple access points (APs), but this will increase
the overhead and infrastructure cost.
As a different approach, Wu et al. developed a single-AP
indoor positioning algorithm that could achieve centimeterlevel accuracy with a single realization of online measurement
by utilizing the TR technique [30]. Since TR is able to focus the
energy of the transmitted signal only onto the intended location,
by utilizing a unique, location-specific CSI, the TRIPS [30]
can position a user by matching the CSI with the geographical
location [37]. Since spatial focusing is a half-wavelength focus
spot, the TRIPS can achieve a 1- to 2-cm level positioning
62

K1 K2

m ,
2

(1)
2

with h = max z R kN=sub1 h F, 1 [k]h F, 2 [k]k* e-jkz , K 1 = R kN=sub1 h F, 1[k] ,
2
and K 2 = R kN= 1 h F, 2 [k] , where h F, 1 and h F, 2 represent two
CSI vectors in the frequency domain [i.e., the channel frequency response (CFR)], N sub stands for the number of usable
subcarriers, h F, 1 [k] and h F, 2 [k] are the CFRs on subcarrier
k, h is the modified cross correlation between h F, 1 and h F, 2
after compensating for the phase distortions caused by synchronization errors, and K 1, K 2 are the CFR energies with
respect to h F, 1 and h F, 2 .
The experimental results in Figure 3 illustrate the TRRS
distribution among the central location on the experimental
structure and its nearby locations under different effective bandwidths. The figure shows that 40 MHz of bandwidth is insufficient to distinguish nearby locations centimeters away. The
ambiguity decreases significantly with an increasing bandwidth.
When the effective bandwidth reaches 360 MHz, the region of
ambiguity shrinks to a ball with an approximately 1-cm radius,
which indicates centimeter-level accuracy.
Unfortunately, the bandwidths on mainstream 802.11n
Wi-Fi chips are merely 20 or 40 MHz, insufficient for centimeter-level indoor positioning. This motivates the formulation of a large effective bandwidth by exploiting the diversities
on Wi-Fi devices, i.e., the frequency and spatial diversity.
Figure 4 shows a generalized framework of diversity
exploitation. More specifically, Figure 4(a) shows an example
of fusing CSI values from four different Wi-Fi channels, while
Figure 4(b) demonstrates merging the CSI values from four
receiving antennas. Both diversities can be exploited at the
sub

Centimeter-accuracy indoor positioning with Wi-Fi

IEEE Signal Processing Magazine

May 2018

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