LiFi Experiments in a Hospital

Abstract

We present LiFi channel measurements in a neurosurgery room of Motol University Hospital in Prague. Individual channels are combined into a virtual multiuser MIMO link. We report achievable data rates for different LiFi transmission schemes.

Full text

LiFi Experiments in a Hospital
Sreelal Maravanchery Mana1, Peter Hellwig1, Jonas Hilt1, Kai Lennert Bober1, Volker Jungnickel1
Klara Hirmanova2, Petr Chvojka3, Radek Janca3, Stanislav Zvanovec3
1Fraunhofer Institute for Telecommunications, Heinrich-Hertz Institute (HHI), Einsteinufer 37, 10587 Berlin, Germany
2Department of Medical Technology, Motol University Hospital, Prague 150 06, Czech Republic
3Faculty of Electrical Engineering, Czech Technical University in Prague, Prague 16627, Czech Republic
sreelal.maravanchery@hhi.fraunhofer.de
Abstract: We present LiFi channel measurements in a neurosurgery room of Motol University
Hospital in Prague. Individual channels are combined into a virtual multiuser MIMO link. We
report achievable data rates for different LiFi transmission schemes.
OCIS codes: (060.0060) Fiber optics and optical communications, (060.2605) Free-space optical communication
1. Introduction
The connected or smart hospital is widely considered as the future of healthcare, where all medical devices are
connected to a wireless infrastructure and provide accurate and timely information [1-4]. Aiming at safe and secure
wireless communication, the use of wireless local area networks such as WiFi in surgery environments is limited
due to possible electromagnetic interference (EMI). Hence, the deployment of a networked optical wireless
technology (LiFi), based on the light has the potential to be employed in hospital environments [2]. Recently,
several research groups provided insights into the potential of LiFi in healthcare environments [1-3]. For introducing
LiFi in hospitals, engineers need a fundamental understanding of the propagation scenario, i.e. the impact of
blockages and reflections by objects. Accordingly, robust transmission schemes need to be developed to enable
reliable wireless communication and high data rates. In this paper, we report for the first time on an experimental
investigation of LiFi transmission in a typical neurosurgery room. We recorded frequency responses of the optical
channels for various transmitter (Tx) and receiver (Rx) positions in the room and considered them subsequently as a
virtual multiuser multiple-input multiple-output (MU-MIMO) system. Finally, the achievable data throughputs for
possible LiFi transmission schemes are estimated.
2. Measurement Scenario
Fig. 1 shows the measurement scenario. Experiments were performed in the children neurosurgery room of Motol
University Hospital in Prague, Czech Republic. Fig. 1(a) shows the photo of the operating room, the positions of Rx
and Tx are highlighted in Figs. 1(b) and (c), respectively. The existing communication systems in this environment
are: i) body sensor network (BSN); ii) infrared (IR) navigation for precise position detection; iii) high-resolution
cameras for monitoring and recording the surgery and iv) endoscopes. Here, we investigate the performance of a
LiFi system employed in the operating room for downlink transmission of BSN. Tx and Rx are each fixed at six
different positions. All Rxs are placed on the operating table with 1 m height looking upwards to the ceiling where
the Txs are mounted at around 2 m height with a certain tilt looking downwards to the operating table. Multiple Rxs
represent multiple sensors, while multiple Txs are used in order to avoid possible blockage of the line-of-sight
(LOS) due to devices and humans moving inside the room.
Fig.1: Measurement Scenario: (a) Neurosurgical operating room; (b) Positions of Rx in the operating table; (c) Top view of the surgery room
where the Tx positions are marked in green color. Here Rx positions are marked in red

All measurements are conducted in the same static environment one after the other for each Tx-Rx link. Here, the
conventional frequency sweep technique is adopted [5]. A standard Agilent vector network analyzer (VNA
E5061B3L5) is connected between Tx and Rx feeding a continuous wave signal with variable frequency from 1 to
200 MHz. For characterizing each channel, the optical frontend response is calibrated out, which is measured at a
minimum distortion-free distance in free LOS [6]. All the channel responses and channel parameters are calculated
as given in [6].
3. Channel Characterization
From the experimental data, it is observed that the signal strength is weaker with respect to all Rx when using Tx1
and Tx2. Tx1 is often blocked by a monitor while Tx2 is too far away from the operating table. Here, we report only
channel characteristics of the other four Tx positions with respect to each Rx position. The measured magnitude
responses of the channels are shown in Fig. 2(a).
Rx1 and Rx2 are placed at the head position of the patient which has good visibility and free LOS from Tx5 and
Tx3. Similarly, Rx4 and Rx5 are nearby the leg position of the patient, where Tx4 and Tx6 have good signal
strength. The amplitude responses for Rx3 and Rx6 are always between the maximum and minimum response, since
they are located at the center position of the operating table. The initial drop at the lower frequencies of the
amplitude response indicates that there are many diffuse reflections in the surgery room where all objects have
bright colors (see Fig. 1(a)). Flatness at higher frequencies is due to the presence of the LOS.
4. MU-MIMO scenario
In our architecture, we assume that all transmitters are coordinated from a central unit which enables simultaneous
transmissions from multiple transmitters, i.e. spatial diversity to prevent sudden blockages of the LOS. Moreover,
joint scheduling of multiple user transmissions is possible in the space and time domains. The MU-MIMO link is
commonly characterized by using the singular values of the compound channel matrix H, which contains all the
links between all Txs and all Rxs. The singular values of H are plotted in Fig. 2(b) which shows that 2 of them are
much stronger than the two others. This suggest that at least 2 data streams can be spatially multiplexed, i.e.
transmitted in parallel, in our specific scenario. Figs. 2(c) shows the normalized channel gains at 1 MHz between all
Txs and all Rxs, and indicates that each Rx has at least 2 strong Tx links. Fig. 2(d) confirms our expectation that the
delay spread is low if the path gain is high and vice versa. From our observations, it is intuitive that Tx3 and Tx5
send a first data stream to Rx1 or Rx2. At the same time, Tx4 and Tx6 send a second data stream to Rx3, Rx4, Rx5
or Rx6. In this way, each Rx has data signals coming from two Txs and potential interference from two other Txs.
This ensures reliable transmission as one LOS could be blocked while the transmission is ongoing.
5. MU-MIMO results
Now we assume three possible transmission schemes and evaluate the achievable throughput. By assuming equal
transmit power, the signal to interference noise ratio (SINR) is calculated for each Rx by using the magnitude
response of the channel. For estimation of SINR, however, noise power is unknown. The maximum SNR that can be
achieved using our frontend design is 45 dB. Hence, we assumed a constant noise floor at a level which is 45 dB
Fig. 2: Results: (a) Magnitude response of channels for each Tx; (b) Singular values of 4x6 distributed MIMO; (c) Normalized gain at 1
MHz. Here first maxima indicated as X and second maxima as O; (d) Normalized delay spread of the channel

below the maximum magnitude response of the LiFi system measured at the minimum distortion-less distance. From
the estimated SINR, the throughput TP of each individual Rx is calculated by means of a modified Shannon formula
TP=B*log2(1+SINR/

) where B is the bandwidth and

=10 is assumed taking imperfect constellation shaping,
limited code word length and limited modulation index to avoid clipping into account. Note that the above formula
is applied on each frequency and then summation is applied over the entire bandwidth, i.e. 200 MHz.
The first scheme is classical time-division multiple access (TDMA) where each user is assigned one out of six
exclusive time slots and served by its best two Txs. In the second scheme, TDMA is extended by space division
multiple access (SDMA) where half of the time in a particular time slot is scheduled for Rx1 and Rx2 in order to
receive first data stream from Tx3 and Tx5 and other Rx has simultaneously one by fourth of the time to receive the
signals from Tx4 and Tx6. Thus, an interference term is included in the SINR calculation depending on the selected
Txs and Rx. Finally, the zero forcing (ZF) method has been applied after the same scheduling used in the TDMA-
SDMA approach. Note that the effective MIMO channel gain matrix in each time slot is given as 2x2 matrix. A
Moore-Penrose pseudo-inverse of the channel gain matrix is calculated as in [7]. All elements of the inverse matrix
are normalized with respect to the row norm to avoid the power fluctuations and send with equal power to each user.
Here the ZF is computed in each time slot and it executes the channel inversion for removing the interference. Note
that, Rx1 and Rx2 have higher throughput since they are available for longer time with stronger signals.
Fig. 3 shows the estimated throughput of the investigated configurations. From Fig. 3(a) it is clear that the ZF
method brings a significant improvement in the throughput for each Rx. For Rx3 and Rx6, TDMA method is much
better than SDMA and TDMA due to high interference from the other sources. The total throughput of the system is
shown in Fig. 3(b). We can observe that the throughput of the system with ZF is around two times higher than
TDMA alone and 1.5 times higher compared to the interference limited case.
5. Conclusion
In this paper, we report the first experimental study of LiFi system in a medical scenario. The frequency response of
the LiFi channels were measured in a neurosurgical operating room and corresponding channel parameters were
calculated. By considering all measured channels as a virtual distributed MU-MIMO system, multiplexing schemes
such as i) TDMA; ii) TDMA-SDMA and iii) TDMA-SDMA with ZF were considered and the achievable
throughputs were estimated. We observed that TDMA-SDMA with ZF provides the highest possible throughput.
6. Acknowledgement
The authors acknowledge the Department of Neurosurgery, 2nd Faculty of Medicine, Charles University in Prague
and Motol University Hospital, Prague, Czech Republic for all the provided help during experiments. This research
is funded by the VisIoN, a European Union’s H-2020 MSCA ITN program under the grant agreement no 764461.
7. References
[1] Y.K. Cheong et al., “Hazardless biomedical sensing data transmission using VLC,” IEEE Sensors J., vol. 13, no. 9,pp. 3347–3348, Sep. 2013.
[2] L. Chevalier et al., "Optical Wireless Links as an Alternative to Radio-Frequency for Medical Body Area Networks," JSAC, vol.33, no.9,
pp.2002-2010, Sept. 2015.
[3] S. S. Torkestani et al., “Indoor optical wireless system dedicated to healthcare application in a hospital,” IET Commun., vol. 6, no. 5, pp. 541–
547, Mar. 27, 2012.
[4] R. Murai et al., "A novel visible light communication system for enhanced control of autonomous delivery robots in a hospital," SII,
IEEE/SICE International Symposium on, pp. 510-516, Fukuoka, 16-18 Dec 2012.
[5] J. M. Kahn et al., "Experimental characterization of non-directed indoor infrared channels," in IEEE Transactions on Communications, vol.
43, no. 2/3/4, pp. 1613-1623, 1995.
[6] S. M. Mana et al., "Experiments in Non-Line-of-Sight Li-Fi Channels," 2019 Global LIFI Congress (GLC), Paris, France, 2019, pp. 1-6.
[7] Q. H. Spencer et al., “Zero-forcing methods for downlink spatial multiplexing in multiuser MIMO channels,” IEEE Trans. Signal Proc ess.,
vol. 52, no. 2, pp. 461–471, 2004.
Fig. 3 Results: (a) Throughput of each Rx at 10 MHz; (b) Total throughput of the system at different transmission schemes at 10 MHz