994 IEEE COMMUNICATIONS LETTERS, VOL. 22, NO. 5, MAY 2018
Efficient Visible Light Sensing in Eigenspace
Shuping Hu, Qian Gao , Chen Gong , and Zhengyuan Xu
Abstract—Due to the wide deployment of light-emitting diode
for indoor illumination, a secure and highly sensitive visible light
sensing (VLS) technology can be adopted for situation awareness
in smart building management and channel update in visible-light
communication. Based on existing lighting devices, a device-free
VLS system with the photodetectors on the ceiling is proposed.
Since any kind of activity may lead to the change of indoor
channel matrix, a real-time window matrix is constructed to sense
the indoor activity, and random matrix theory is employed, where
the likelihood-ratio test (LRT) and mean spectral radius (MSR)
serve as the system variation indicators. We compare the average
sensing error rates of LRT detection and MSR detection with
that of average intensity variation detection on people movement,
which show the performance gain of the proposed eigenspace
detection approaches. An efficient VLS can also be used to
balance the communication performance and channel update
overhead.
Index Terms— Visible light sensing, likelihood-ratio test, mean
spectral radius, eigenspace.
I. INTRODUCTION
W
ITH the wide deployment of light-emitting diodes
(LEDs) for illumination, visible light communica-
tion (VLC) has become a viable candidate for indoor
communication [1]. Recently, visible light sensing (VLS) has
been considered because VLS can use the high-power LEDs
and highly sensitive photodiodes (PDs) in VLC, which implies
that the LEDs can be used for illumination, communication
and sensing simultaneously. VLS can help to realize situa-
tion awareness (SA) for broad applications in smart building
management [2], such as fire warning, antitheft, and occupancy
inference, the channel update in VLC can be achieved based
on VLS as well. These make VLS a strong candidate for
sensing as infrared (IR) sensing [3] and acoustic/ultrasonic [4].
Since the wavelength of visible light is nanoscale, visible light
cannot penetrate walls, such that VLS can preserve privacy
unlike the other solutions such as WiFi [5].
Different from the traditional multi-LEDs’ intensity based
indoor objects’ location scheme that requires the users to
carry a sensing device [6], it is more interesting to develop
techniques free of sensing devices at users [7]–[9], to sense
the change of indoor environment, which is more pragmatic in
the applications in VLS that do not need to have the objects’
position information. The light sensors can be installed on
Manuscript received January 18, 2018; accepted February 16, 2018. Date of
publication February 22, 2018; date of current version May 8, 2018. This work
was supported by Key Program of National Natural Science Foundation of
China (Grant No. 61631018), National Natural Science Foundation of China
(Grant No. 61501420), Key Research Program of Frontier Sciences of CAS
(Grant No. QYZDY-SSW-JSC003), and the Fundamental Research Funds for
the Central Universities. The associate editor coordinating the review of this
paper and approving it for publication was L. Mucchi. (Corresponding author:
Zhengyuan Xu.)
The authors are with the Key Laboratory of Wireless-Optical
Communications, Chinese Academy of Sciences, University of Science
and Technology of China, Hefei 230027, China (e-mail: xuzy@ustc.edu.cn).
Digital Object Identifier 10.1109/LCOMM.2018.2808498
TABLE I
T
YPICAL PARAMETERS OF INDOOR ENVIRONMENT
the floor [7] or on the ceiling [8], [9]. Average intensity
variation (AIV) detection is adopted to sense the change of
indoor environment in existing works. However, AIV detection
may suffer high sensing error rate.
In this letter, we propose two novel detection approaches in
eigenspace to conduct VLS, which outperform AIV detection.
Given the emitted intensity, the received data of the ceiling
photosensors is random due to the noise. A real-time window
matrix is adopted to conduct real-time VLS based on random
matrix theory (RMT) [10]. Likelihood-ratio test (LRT) detec-
tion based on Marchenko-Pastur law (M-P law) and mean
spectral radius (MSR) detection based on ring law are adopted
to detect the change of indoor environment such as people
movement. The average sensing error rates of LRT detection
and MSR detection are lower than that of AIV detection
especially in the low SNR region. VLS can also be used
to balance the bit error rate (BER) performance and channel
update overhead in VLC.
II. C
HANNEL MODEL
In order to expand the sensing area and increase the sensing
accuracy, multiple LEDs are deployed as the transmitter
and multiple PDs are deployed as the receiver, both on
the ceiling. Similar to VLC, we can build the relationship
between the LED emitting intensity and PD receiving intensity
as a multiple-input multiple-output communication (MIMO)
system, shown as follows,
y = Hx + n, (1)
where n is the thermal noise and n ∼ N
0,σ
2
n
I
. Due to the
constant indoor illuminary level, the emitted intensity vector
x is constant. Since any activity of the indoor objectives may
lead to the change of indoor channel matrix and the detected
intensity, we can develop an efficient algorithm to infer the
change of channel matrix H based on the received signal
intensity y.
We first obtain the channel impulse response (CIR) of each
PD [11] to estimate the channel matrix H using Zemax,
a commercial optical simulation and design platform. For
example, in an empty room with the typical parameters shown
in Table I, we consider the situation that one person stands
in different positions, shown in Figure 1(a) and Figure 1(b).
We can obtain channel matrix H under different indoor
1558-2558 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.