IEEE Consumer Electronics Magazine - March/April 2021 - 93

Keeping this in mind, a passive system, called
CommSense (communication-based environment
sensing system), is proposed here for measuring
crowd size in an outdoor environment. The concept of CommSense was proposed and verified by
the authors in simulation and with field-collected
data for various indoor applications, such as
indoor object detection, indoor localization, and
indoor occupancy estimation.6 The system was
also tested for outdoor vehicle detection and classification.7 These give the confidence to apply the
CommSense principle. A major advantage of the
CommSense system, compared to audio or
vision-based methods, is that our system is nonintrusive. It cannot identify the persons.
The experiments and analyses performed
here are as follows.
1) First, we considered that the people in the
outdoor crowd are static. Crowd detection
using a threshold-based method was performed first. If a crowd is detected, it triggers
the investigation of the deriving number of
persons in the crowd.
2) In the next step, the same analysis was
repeated when the people in the crowd were
moving freely.
3) Finally, we have performed these two experiments on a different day and time to examine
the consistency of this approach.

STATE OF THE ART
As crowd sensing is vital for many other applications,8,9 there have been attempts to perform
outdoor-crowd-size estimation in the past. Videobased crowd-size estimation was proposed by Li
et al.10 The performance of video processing for
crowd-monitoring applications was analyzed by
Bailas et al.11 A crowd-counting method was proposed by Xing et al.,12 which uses detection flow
along the temporal video sequence. Automated
video analytic was used for crowd monitoring and
counting in the work by Cheong et al.13 Counting
the number of people present in a crowd with a
real-time network of image sensors was proposed
by Yang et al.14. A motion-based crowd density
estimation method was proposed by Chondro
et al.15 A crowd-counting method based on image
processing and convolutional neural network was

March/April 2021

proposed by Liu et al.16 But these vision-based
approaches require the crowd to be in the line-ofsight of the cameras. Their performance also
depends on the overlapping of the objects, the relative positioning of the crowd, visibility at the time
of the day, weather conditions, pollution level, etc.
Additionally, it also poses privacy concerns. Danielis et al.17 tried to estimate the size of dense
crowds, using a distributed protocol that relies on
mobile device-to-device communication. This
required active transmission at all times and was
verified in simulation. In addition, we need crowdsize estimation on a smaller scale. Khan et al.18 use
smartphones' acoustic sensors in the presence of
human conversation, and motion sensors in the
absence of any conversational data, for crowd-size
estimation. There has been few attempts to use
LTE-based systems for crowd-density estimation.19,20 Tripathy et al.21 presented an Internet-ofMedical-Things-enabled wearable called EasyBand
for autocontact tracing. Currently, the world is
almost coming to a halt to reduce COVID-19
spread. Official recommendations for social distancing have pushed people into ever-smaller
clusters.1

LTE-CommSense SYSTEM
Figure 1(a) shows a simple version of the proposed system that uses a piece of single user
equipment (UE).6,7 In this article, a single UE
was considered for verification using practical
data. The LTE UE receiver works at Band-40
(2300-2400-MHz frequency band) in time division duplexing topology. The signal captured
had a bandwidth of 10 MHz.
In the experimental setup, we have used a total
of 19 number of people in an outdoor roadside
environment. The CommSense prototype was
employed to collect LTE downlink (DL) data
affected by the presence of the crowd. The modeled UE, which is a part of the proposed prototype,
then receives the DL data and performs standard
UE operations to evaluate state information (CSI).
The description and working principle of an
USRP N200 SDR platform can be obtained from
the work by Sardar et al.6,7 The SDR was modeled
as the CommSense prototype containing LTE
receiver. The LTE DL signal was captured via the
antenna and radio frequency (RF) daughter card

IEEE Consumer Electronics Magazine - March/April 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March/April 2021